Who Makes Acquisitions?

A Test of the Overconfidence Hypothesis ∗

Ulrike Malmendier

Stanford University

ulrikem@stanford.edu

Geoffrey Tate

University of Pennsylvania

tate@wharton.upenn.edu

September 28, 2004

Abstract

We analyze whether the volume and returns of merger activities are affected by CEO

overconfidence. Overconfident CEOs over-estimate their ability to generate returns and

perceive outside finance to be over-priced. As a result, they undertake value-destroying

mergers when they have abundant internal funds, and they may forego value-creating merger

opportunities when they need to raise external funds. In order to test the overconfidence

hypothesis, we employ detailed data on personal portfolio decisions of CEOs in Forbes

500 companies. We identify CEOs who, despite their under-diversification, systematically

fail to reduce their personal exposure to company risk by exercising highly in the money

stock options. We find that these CEOs do not profit from holding their options relative

to an alternative diversification strategy. In addition, we find that such CEOs are more

acquisitive on average and particularly prone to undertake diversifying deals. As predicted

by the overconfidence hypothesis, the effects are largest in firms with abundant cash and

untapped debt capacity. The same results hold for CEOs whom the press describes as

“confident” or “optimistic.” Moreover, the market reacts significantly more negatively to

takeover bids by such CEOs. (JEL G34, G14, G32, D80).

“Many managements apparently were overexposed in impressionable childhood

years to the story in which the imprisoned handsome prince is released from a toad’s

body by a kiss from a beautiful princess. Consequently, they are certain their man-

agerial kiss will do wonders for the profitability of Company T[arget]...We’ve ob-

served many kisses but very few miracles. Nevertheless, many managerial princesses

remain serenely confident about the future potency of their kisses-even after their

corporate backyards are knee-deep in unresponsive toads.”

-Warren Buffet, Berkshire Hathaway Inc. Annual Report, 19811

Mergers and acquisitions are among the most significant events to occur in corporations, giving

rise to massive reallocation of financial and human capital. The staggering economic magnitude

of these deals has inspired a myriad of research on their causes and consequences. Most theories

focus on the efficiency gains that motivate takeover activities. The empirical results on returns

to mergers, however, are mixed. Mergers appear to have destroyed value on average in the

merger wave of the late nineties; but they may have created value during earlier periods.2

Moreover, even in the case of value-creating mergers, the gains typically do not accrue to the

shareholders of the acquiring company. Rather, there is a significant positive gain in target

value upon the announcement of a bid, and a significant loss to the acquiror.3 These findings

suggest that mergers are often not in the interest of the shareholders of the acquiring company.

Overconfidence among chief executive officers (CEOs) has long had popular appeal as an

explanation for these findings.4 Overconfident CEOs overestimate their leadership skills and

ability to generate returns. As a result they may engage in mergers that are value-destroying

for their shareholders. Roll (1986) first formalized this notion, linking it to the winner’s curse

among bidders.5 In this paper, we develop a model of CEO overconfidence that accounts for

the difference in beliefs between CEOs and the outside market. We then provide empirical

evidence that CEOs who systematically overestimate the future returns of their company tend

1

to do more mergers and mergers that are not likely to generate value, particularly when they

have access to abundant internal financing.

The overconfidence hypothesis builds on the psychology literature on self- enhancement, which

demonstrates that individuals tend to overestimate their abilities relative to an anonymous

benchmark or to their peers.6 The “better than average effect” also affects the attribution of

causality. Because individuals expect their behavior to produce success, they attribute good

outcomes to their actions, but bad outcomes to chance (Miller and Ross, 1975). This self-

serving attribution of outcomes reinforces individual overconfidence.7 According to a number

of studies, executives appear to be particularly prone to display overconfidence (Kidd, 1970;

Larwood and Whittaker, 1977; Moore, 1977; Kahneman and Lovallo, 1993). Baron (2000)

surveys related literature on “cognitive factors in entrepreneurship,” noting prominently the

tendency of entrepreneurs to be overconfident in their own judgements. Subsequent studies

have found experimental evidence on overconfidence in market entry decisions (Camerer and

Lovallo, 1999) and on the underestimation of cultural conflicts in mergers (Weber and Camerer,

2003).

Merger decisions are an ideal setting to test for the effects of individual overconfidence. First,

mergers are among the most significant corporate decisions and, as a result, require the di-

rect oversight of the CEO. Second, while not all decision-making is likely to be affected by

overconfidence, merger decisions are exactly the kind of choice psychologists link to overconfi-

dence. Namely, they argue that individuals are especially overconfident about outcomes that

they believe are under their control and about outcomes to which they are highly committed

(Weinstein, 1980; Weinstein and Klein, 2002). Both scenarios apply to the merger decisions

of CEOs. A CEO who conducts a merger is ostensibly replacing the current management of

the target firm with himself. Therefore, he is likely to feel the illusion of control over the

outcome and to underestimate the likelihood of eventual failure (Langer, 1975; March and

2

Shapira 1987). Moreover, a successful merger enhances the CEO’s professional standing and

his future employment prospects. In addition, the typical compensation contract of a CEO ties

his personal wealth to the company’s stock price and, hence, to the outcomes of his acquisition

decisions. Finally, overconfidence is considerably more likely to affect market prices and the

wealth of other market participants in the merger context than, for example, in the context

of stock trades by individual investors. Arbitrage against an overconfident CEO may requires

removal of the CEO from power, which is likely to be difficult and costly. As we will see

in Section IV, the market does indeed react more negatively to the merger announcement of

CEOs whom we classify as overconfident. However, this rarely stops CEOs from implementing

the merger. In other words, while financial markets may adjust prices to reflect the level of

value destruction, this does not prevent the CEO from taking the value-destroying action.

Our model of overconfidence builds on the assumption that overconfident CEOs overestimate

their ability to select profitable future projects, whether in their current company or in a

merged company. They may also overestimate the synergies between their company and a

potential target, or underestimate how disruptive a merger will be. As a result, overconfidence

may induce mergers that are, on the margin, value-destroying. Whether an overconfident

CEO undertakes a merger depends, however, also on the availability of internal financing.

Overconfident CEOs view their company as undervalued by outside investors who are less

optimistic about the prospects of the firm. This perceived undervaluation makes overconfident

CEOs reluctant to issue equity and they may thus forego value-creating merger opportunities

when they need to raise external funds to finance the merger.

The trade-off between (perceived) undervaluation and (perceived) high returns from acquisi-

tions leaves the question of whether overconfident CEOs are more likely, on average, to conduct

mergers an empirical matter. A positive net effect would underscore the empirical relevance

of overconfidence. However, the model makes the unambiguous prediction that overconfident

3

managers are more likely to conduct value-destroying mergers. Overconfident managers are

also more likely to conduct mergers if they have abundant internal resources and do not need to

issue “undervalued” equity to finance the deal. Moreover, the lower average quality of mergers

undertaken by overconfident CEOs and the tendency of overconfident CEOs to overpay for

their acquisitions in the face of competition should be reflected in a (more) negative market

reaction to the merger announcement.

We then test empirically whether differences in merger activities can be linked to differences

in CEO characteristics and beliefs. Using data on the personal portfolio decisions of CEOs

in Forbes 500 companies, we identify CEOs who systematically overinvest in stock of their

company. Here, we build on previous literature in corporate finance that shows that CEOs

should exercise executive stock options with positive value (beyond a certain threshold) well

before expiration due to the suboptimal concentration of their portfolio in company-specific

risk.8 We identify CEOs who display the opposite behavior, i.e. who fail to diversify company-

specific risk on their private portfolio and systematically hold on to company stock options, as

persistently bullish about their company’s future prospects. We then show that this bullishness

does not appear to reflect insider knowledge since the hypothetical returns CEOs could have

obtained by exercising their options earlier are positive on average. Furthermore, the delay in

option exercise is also unlikely to be a ‘signal’ about the company or merger quality since the

returns to mergers of CEOs who overinvest in their company are more negative than those of

their colleagues with more diversified private portfolios. We thus interpret the overinvestment

of CEO in their own company as an indication that they overestimate the future returns of

their company and classify them as overconfident.

We construct several measures of overconfidence following this logic. First, we consider CEOs

who hold options all the way to expiration. Provided that the options exceed a benchmark

for rational exercise motivated by the Hall and Murphy (2002) model, we classify the CEOs

4

as “Longholders.” While unnecessary to generate our results, the threshold makes it unlikely

that the CEOs’ decisions to hold were driven by an optimal tradeoff of option value and di-

versification. Second, we split the Longholder measure into two pieces: “Pre-Longholder” and

“Post-Longholder.” Post-Longholder identifies only the Longholder CEO years after the first

time the CEO held options to expiration. Pre-Longholder captures the remaining Longholder

CEO years. Finally, we construct an alternative measure, Holder 67, based on the CEO’s exer-

cise decision for options with 5 years remaining duration. We, again, compute an appropriate

threshold for rational exercise using a reasonable calibration of the Hall-Murphy model. When

a CEO has an option with 5 years remaining duration that is beyond the threshold, but fails

to exercise, we set the dummy variable Holder 67 equal to 1 for the remainder of the CEO’s

tenure. CEOs in the same situation who instead exercise have Holder 67 equal to 0.

Linking CEOs private portfolio strategies to their corporate decision-making, we find that

overconfident CEOs are significantly more likely to conduct mergers than rational CEOs at

any point in time. The higher acquisitiveness of overconfident CEOs — even “on average” —

suggests that overconfidence is an important determinant of merger activity. Further, we find

that the heightened merger activities of overconfident CEOs are primarily due to an increased

likelihood of conducting diversifying acquisitions. Previous literature suggests that diversifying

mergers are unlikely to create value in the acquiring firm.9 Thus, it is consistent with our theory

that overconfident managers are particularly likely to undertake them. Second, we find that

the relationship between overconfidence and the likelihood of doing a merger is strongest when

CEOs can avoid equity-financing., i.e. in the least equity dependent firms. Overconfident

CEOs strongly prefer cash- or debt-financed mergers to stock deals unless their firm appears

to be overvalued by the market.

To bolster our portfolio measure of overconfidence, we construct an alternative measure based

on how a CEO is characterized in the press. We analyze the difference in merger activity be-

5

tween CEOs who are portrayed in the business press as “confident” and “optimistic” and CEOs

who are portrayed instead as “reliable,” “cautious,” “conservative,” “practical,” “frugal,” or

“steady.” Since press coverage during mergers may mechanistically lead to more coverage as

“confident” and “optimistic” we employ a hazard-rate type approach and restrict the analysis

to CEO-firm years up to the first merger of a CEO (if any). Controlling for the total number

of press mentions, we then perform the same empirical analyses as with the portfolio overcon-

fidence measures. All results replicate. Furthermore, the measures are highly correlated.

Finally, we look directly at the market’s perception of the merger decisions made by overcon-

fident CEOs. Using standard event study methodology, we show that outside investors react

more negatively to the announcement of merger bids by overconfident CEOs. This result holds

even controlling for relatedness of the target and acquiror, ownership stake of the acquiring

CEO, corporate governance of the acquiror, and method of financing the merger. Our results

suggest that, even if overconfident CEOs create firm value along some dimensions10, mergers

and acquisitions are not among them.

Our theory of managerial overconfidence provides a natural complement to standard agency

theory. Both “empire-building preferences” and overconfidence predict heightened managerial

acquisitiveness — especially given abundant internal resources — and, as shown in Malmendier

and Tate (2003), a heightened sensitivity of corporate investment to cash flow. Unlike empire-

builders, however, overconfident CEOs, believe that they are acting in the interest of the

shareholders. Thus, overconfidence, cast as an agency problem, challenges the effectiveness of

stock and option grants to top executives as an incentive mechanism. On the other hand, it

provides additional underpinning for models of debt overhang. High leverage may effectively

counterbalance an overconfident CEO’s eagerness to invest and acquire, given his reluctance

to issue equity he perceives as undervalued. In addition, the failure of traditional incentives to

mitigate overconfidence underscores the importance of an independent board of directors.

6

This paper contributes to the growing literature in behavioral corporate finance which seeks

to understand the implications of biased decision making in the context of organizations.

One approach has been to analyze how managers exploit biases among investors, resulting in

inefficiencies in the capital market.11 More recently, researchers have started to analyze biases

that affect corporate decision-makers themselves, such as overconfidence or sunk-cost fallacy.12

The promise of the latter approach is that we have more precise knowledge about the types

of biases affecting high-level executives than about the biases relevant for a broad mass of

individual investors. Moreover, the argument for limited arbitrage is particularly simple and

salient in this context.

The paper is organized as follows. In Section I we present a simple model of managerial over-

confidence. In Section II we introduce the data. Section III introduces our empirical measures

of overconfidence, based on option-holding and press-coverage, and discusses alternative inter-

pretations. We then describe the empirical strategy and provide evidence that overconfidence

can explain managerial acquisitiveness. In Section IV, we study the market reaction to mergers

by overconfident CEOs. Section V concludes and provides some broad directions for future

research.

I Theory

We construct a simple model that demonstrates the effects of managerial overconfidence on

merger decisions in an otherwise frictionless market. In particular, we assume symmetric

information between corporate insiders and outside investors. Moreover, management acts in

the interests of current shareholders. We consider first a world with a single bidder for the

target company and assume that the acquiror can extract the full surplus. We then show

how variations in the relative bargaining power of target and acquiror, for example due to

7

competition among bidders, affect the potential over-payment by overconfident acquirors.

A Single Acquiror with Full Bargaining Power

Denote the market value of Acquiror A and Target T as VA and VT respectively. The CEO

of A chooses whether or not to acquire T . He has access to an amount c of internal resources

(cash and riskless debt). We denote the amount paid to the target shareholders as part of the

merger financing as c, with c · c. V (c) is the market value of the combination of A and T ,bV (c) the A manager’s valuation of the combination of A and T , and bVA his perception of hisown company’s value if he does not pursue the merger. An overconfident CEO overestimates

the returns he will generate and thus overvalues his own company, bVA > VA, as well as the

merger, bV (c)− V (c) > bVA − VA for some c.

Since the acquiror has all bargaining power, the manager of A must pay VT for the target,

independent of his degree of overconfidence. If the A manager offers an amount c < VT of cash

financing (or other non-diluting assets), target shareholders demand a share s of the merged

company such that sV (c) = VT − c.

If the A CEO is rational, he chooses to conduct the takeover if and only if V (c)−(VT −c) > VA.

Denoting the merger synergies as e ∈ R , we can decompose V (c) into

(1) V (c) = VA + VT + e− c

Not surprisingly, the rational CEO makes the first best acquisition decision and decides to

acquire whenever e > 0. Moreover, his decision is independent of c. Since the capital market

is fully efficient, there is no extra cost of raising external capital to finance the merger and the

CEO is indifferent among cash, equity, or a combination.

An overconfident CEO overestimates the returns to mergers. Since bV (c) > V (c), the acquiring

CEO also believes that (partial) equity financing entails a loss to current shareholders of (V −cV (c) −

8

V −cbV (c) )bV (c).13 He undertakes the merger despite this perceived cost if he believes the value ofthe diluted shares in the merged company to A’s current shareholders is greater than the value

of A forsaking the merger. That is, he undertakes the merger if and only if (1− s)bV (c) > bVAfor some c · c. Substituting for s, he acquires T iff bV (c)−(VT − c)−

[bV (c)−V (c)](V −c)V (c) > bVA for

some c. Denoting the “perceived” additional merger synergies as be ∈ R++,14 we can decomposebV (c) into(2) bV (c) = bVA + VT + e+ be− c

Then, using (1) and (2), the overconfident manager’s decision rule is to merge whenever e+be >

(bV −V +be)(V −c)V (c) . That is, he merges whenever total perceived merger synergies exceed the

perceived loss due to dilution. Combining these results with the results of the prior section

yields the following lemma and propositions.

Lemma 1 An overconfident CEO exhausts his supply of internal (non-diluting) assets before

issuing equity to finance a merger.

Proof. An overconfident CEO perceives the post-acquisition value of the firm to current share-

holders as G = (1− s)bV (c) = V (c)−V +cV (c)

bV (c) = (V +e)(bV +V +e+be−c)V +V +e−c , where the last equality

uses (1) and (2). Then ∂G∂c =

(V +e)(bV −V +be)(V (c))2

> 0 (as bVA > VA and be > 0 by assumption).

Post-merger value is maximized on c ∈ [0, VT ] by setting c as high as possible. Q.E.D.

Proposition 1 A rational CEO never conducts a value-destroying merger. An overconfident

CEO conducts a value-destroying merger if the perceived synergies e are sufficiently large rel-

ative to the perceived undervaluation (bVA − VA) and the portion of the deal financed by equity

V −cV .

Proof. The first-best decision rule of a rational CEO immediately implies that he does not

conduct a value-destroying merger. An overconfident CEO conducts a merger whenever e+be >

9

(bV −V +be)(V −c)V (c) . Thus, if e · 0, he still conducts the merger as long as be > |e| and bVA − VA

and V −cV (c) are sufficiently small. Q.E.D.

Proposition 2 (i) If c ≥ VT , an overconfident CEO conducts any merger a rational CEO

would conduct and some mergers a rational CEO would not conduct. (ii) If c < VT , an

overconfident CEO does some (value-destroying) mergers a rational CEO would not and a

rational CEO does some (value-creating) mergers that the overconfident CEO would not.

Proof. If c ≥ VT , the overconfident manager sets c = VT by Lemma 1 and be > 0. The resulting

condition for conducting the merger is e + be > 0. Since the rational CEO merges whenever

e > 0, the first part of Proposition 2 follows. For c < VT , the first statement follows from

Proposition 1. To show the second statement in (ii), suppose e > 0. Then, the rational

CEO always does the merger. The overconfident CEO will not do the merger if and only if

e+be <(bV −V +be)(V −c)

V (c) , i.e. if be is sufficiently small and bVA−VA orV −cV (c) are sufficiently large.

Q.E.D.

B Competing Acquirors

We now consider the implications of reduced bargaining strength and less surplus extraction

of the acquiror. Note that, in a standard two-player bargaining framework, the differences in

beliefs between an overconfident A manager and a rational T manager about the size of the

surplus from merging would require assumptions not only about relative bargaining power but

also about higher-order beliefs regarding the parties’ perception of surplus and the interaction

of these beliefs with the parties’ relative bargaining strength. For simplicity, we will thus

endogenize variations in the amount of surplus A can extract by introducing competition

among potential acquirors in a bidding framework.

Suppose that there are I potential acquirors Ai, i = 1, ..., I. Denote by Wi the Ai man-

10

ager’s maximal willingness to pay for T. Wi is simply the market value of the target plus the

(perceived) surplus to Ai’s current shareholders as a result of the merger, i. e.

1. Wi = VT + ei if the Ai manager is rational;

2. Wi = VT + ei + bei − 1{c <W }(bV −V +be )(W −c )

V +V +e −c if the Ai manager is overconfident.

In English auction among bidders with maxWi ≥ VT the equilibrium outcome is as follows15:

1. The winning bidder is Ai∗ , where i∗ = argmaxiWi.

2. The winning bid is b∗ = max{(maxi6=i∗ Wi), VT}.

Note that, contrary to Roll’s theory, an overconfident bidder does not always bid higher than

a rational bidder, even if the actual synergies of the merger are smaller for the rational bidder.

In particular, an overconfident bidder who is considerably more overconfident about the value

of his own company than about the merger may lose the takeover contest. Most importantly,

heterogeneity in the merger synergies can increase the transfer to target shareholders and,

when interacted with overconfidence, can lead to over-payment. Denoting the company that

wins the takeover contest as Ai∗ and defining ‘overpayment’ as a transfer from Ai∗ to T that is

higher than the sum of target value and synergies, VT + ei∗ , we have the following proposition.

Proposition 3 If the manager of the winning acquiror Ai∗ is overconfident, he will over-pay

if maxi6=i∗ Wi ∈ (VT + ei∗ , Wi∗).

C Extensions

Before turning to the empirical predictions of the model, we briefly discuss two important

extensions. First, overconfident CEOs might not only overvalue their potential leadership in

other companies, but also the returns from their internal investment projects (Malmendier and

Tate, 2003). This effect could counteract their increased acquisitiveness if resources are scarce.

An extended model of corporate decision-making would include the menus of both potential

11

acquisitions and internal projects. When new resources become available to the CEO, he would

initiate the next project on either or both menus. While relative returns would determine which

project is chosen first, we would expect overconfidence affect to increase the number and type

of projects of both types over time.16

Second, we have focused exclusively on overconfidence in acquiring managers. Indeed, overcon-

fidence may be an important force in distinguishing acquirors from targets. However, target

managers could be overconfident as well. While overconfidence of target managers will not

change the qualitative predictions of our model, it yields many interesting comparative statics.

For example, acquisitions of target firms with overconfident management are more likely to

be hostile takeovers. The overconfident target management might believe they can create at

least as much value as the potential acquirors and, hence, reject shareholder-value increasing

bids as too low. Similarly, we would expect acquirors to pay a higher premium for targets

with overconfident managers, even in friendly deals. As a result, the acquirors of firms with

overconfident managers are likely to be among the most overconfident managers. In both cases,

overconfidence on the side of the target management can be beneficial to the target sharehold-

ers. Unfortunately, we cannot test any of these implications due to data limitations.17

D Empirical Predictions

In the remainder of the paper, we test the empirical implications of our model. To facili-

tate the translation of the model into predictions about a cross-section of CEOs, we suppose

that e is drawn independently from the same distribution for all potential mergers. That is,

overconfident and rational CEOs do not have systematically different merger opportunities.

The first quantity of interest is the difference in the average probability of conducting a merger

between overconfident and rational CEOs. As noted above, overconfidence implies both over-

12

estimation of merger returns and reluctance to raise outside financing to implement a merger.

The net effect of overconfidence on merger frequency is thus ambiguous. However, whether or

not overconfident CEOs are more acquisitive is a key indicator of the importance of overconfi-

dence as a general explanation of observed merger activity. Moreover, the model delivers three

testable predictions. Proposition 1 and Proposition 2 imply (respectively):

Prediction 1. Overconfident CEOs are more likely to conduct mergers that ex ante have a

high probability of failure (and negative expected return).

Prediction 2. Among CEOs with abundant internal resources (e.g. large cash reserves and

low leverage), overconfident CEOs are more likely to conduct acquisitions.

Finally, Proposition 1 and Proposition 2 together imply that mergers conducted by overconfi-

dent CEOs will be worse on average than mergers conducted by rational CEOs. In addition,

Proposition 3 shows that overconfident managers are prone to overpay for their acquisitions.

Since we have maintained the assumption that the market is efficient, all information about

the quality and terms of the deal will be incorporated at the announcement date and we have

the following prediction.

Prediction 3. The difference between the average stock price reaction to the announcement

of a merger bid by an overconfident CEO and the average stock price reaction for a rational

CEO is negative.

Note that the assumption of symmetric information implies that the merger announcement does

not convey any information about the fundamentals of the acquiring company. In practice,

information revelation will have an impact on the announcement effect (e. g. in Hietala et al.,

2002). For simplicity, we assume that the average effect of such information revelation is the

same among overconfident and rational CEOs.

13

II Data

We analyze a sample of 477 large publicly-traded United States firms from the years 1980 to

1994. The core of the data set is described in detail in Hall and Liebman (1998) and Yermack

(1995). To be included in the sample, a firm must appear at least four times on one of the lists

of largest US companies compiled by Forbes magazine in the period from 1984 to 1994.18 The

virtue of this data is that it provides us with detailed information on the stock ownership and

set of option packages — including exercise price, remaining duration, and number of underlying

shares — for the CEO of each company in each year. From this data we obtain a fairly detailed

picture of the CEO’s portfolio rebalancing over his tenure.

We also collect data on how the press portrays each of the CEOs during the sample period. We

search for articles referring to the CEOs in The New York Times, Business Week, Financial

Times, and The Economist using LexisNexis and for articles in the The Wall Street Journal

using Factiva.com. For each CEO and sample year, we record four statistics: the total number

of articles; the number of articles containing the words “confident” or “confidence;” the number

of articles containing the words “optimistic” or “optimism;” and the number of articles con-

taining the words “reliable,” “cautious,” “conservative,” “practical,” “frugal,” or “steady.” We

hand-check each article to be sure that the terms are used to describe the CEO in question. In

the process of scanning the search output, we separate out any articles specifically describing

the CEO as “not confident” or “not optimistic.”

We supplement this CEO-level data with mergers data from the SDC and CRSP merger

databases. Both data sets give us the announcement date and means of financing for mergers

conducted by our sample of firms. The CRSP data set covers only mergers with CRSP-listed

target firms. We use the SDC data to supplement the set of mergers with acquisitions of private

firms, large subsidiaries, and foreign companies.19 We require that the acquiring company

14

acquire at least 51% of the shares of the target (and, hence, control) in the transaction.

Similarly, we omit acquisitions of companies where the acquiror already holds at least 51%

of the shares. Finally, following Morck et al., (1990), we omit mergers in which the value of

the target is less than five percent of the value of the acquiror.20 For most of the paper, we

consider only completed merger bids; however, when we consider market reaction, we include

all merger bids in the estimations.

We supplement the data with various items from the COMPUSTAT database. We measure

firm size as the natural logarithm of assets (item 6) at the beginning of the year. We measure

investment as capital expenditures (item 128), cash flow as earnings before extraordinary items

(item 18) plus depreciation (item 14), and capital as property, plants and equipment (item 8).

We normalize investment and cash flow with beginning of the year capital. Given that our

sample is not limited to manufacturing firms (though it mainly consists of large, nonfinancial

firms), we check the robustness of our results to normalization by assets (item 6). We measure

Q as the ratio of market value of assets to book value of assets. Market value of assets is

defined as total assets (item6) plus market equity minus book equity. Market equity is defined

as common shares outstanding (item 25) times fiscal year closing price (item 199). Book equity

is calculated as stockholders’ equity (item 216) [or the first available of common equity (item 60)

plus preferred stock par value (item 130) or total assets (item 6) minus total liabilities (item

181)] minus preferred stock liquidating value (item 10) [or the first available of redemption

value (item 56) or par value (item 130)] plus balance sheet deferred taxes and investment tax

credit (item 35) when available minus post retirement assets (item 336) when available. Book

value of assets is total assets (item 6).21 Further, we use fiscal year closing prices (item 199)

adjusted for stock splits (item 27) to calculate annual stock returns. We also use CRSP to

gather stock prices and 2 and 4 digit SIC codes for the companies in our sample and the target

firms in CRSP acquisitions. Missing accounting data (largely from financial firms) leaves us

15

with a final sample of 394 firms. As in Malmendier and Tate (2003), we trim cash flow at the

1% level to ensure that our results are not driven by several extreme outliers. However, all

results of the paper can be replicated with the full data set. The outliers only influence the

estimates at all in the regressions on quintiles of the data set in Subsection F and only in the

(interior) quintiles of lesser interest.

In addition, we collected personal information about the CEOs in our sample using Dun and

Bradstreet and Who’s Who in Finance and Industry. We broadly classify a CEO’s professional

background as financial, engineering or miscellaneous. We classify CEOs as having a finance

background if they previously worked in a financial institution or as CFO, treasurer, accountant

or in another finance related position. CEOs have an engineering background if they are

individual patent-holder or previously worked as engineer, in the natural sciences, or in another

technically-oriented position.

Table 1 presents summary statistics of the data. Panel A presents firm-specific variables and

Panel B CEO-specific variables, both for the full set of CEOs and for the subset of CEOs whom

we classify as overconfident based on their option-exercise behavior (Longholder; see the next

section). The mean, median and standard deviation of all variables are remarkably similar for

overconfident and non-overconfident CEOs; only the number of vested options that have not

been exercised is considerably higher among overconfident CEOs. This difference could stem

from overconfidence, as we will see later, but, regardless, we will control for the level of vested

options in all of our regressions. Panel C presents the summary statistics of the CEOs’ press

coverage. While the mean and median number of press mentions in any of the selected business

press outlays are very high (91.8 and 39 respectively), the average number of mentions with

the attributes “confident” or “optimistic” or any of “reliable, cautious, conservative, practical,

steady, frugal” is below 1 and the median is zero. In our empirical analysis we will thus consider

not only the (lagged) number of mentions but also a form of “moving average” (in indicator

16

form) of mentions of each kind over the sample tenure up to the previous year for each CEO.

Panel D presents summary statistics of the mergers undertaken by CEOs in our sample.

III The Impact of Overconfidence on Acquisitiveness

A Measures of Overconfidence

We first analyze private portfolio decisions of CEOs to identify overestimation of future returns.

In particular, we exploit the panel data information on the timing of executive option exercises.

Compensation contracts often grant stock options to the CEO. These options are non-tradeable

and give the executive the right to purchase shares of company stock, usually at the stock price

on the grant date. Most executive options have a ten year life span and are exercisable after (at

most) a four-year vesting period. Upon exercise, the CEO receives shares of company stock;

however, these shares are almost always immediately sold (Ofek and Yermack 2000). Thus,

at exercise the executive receives for each option the value in cash given by the current stock

price minus the strike price (i.e. the stock price on the grant date for options granted “at the

money”).

Black and Scholes (1973) argue that investors should value traded options as if they were risk-

neutral, since they can offset the idiosyncratic risk of any particular company by diversifying

their portfolio. So, since an option is always worth more alive than dead (as option value is

non-negative), they should never exercise options early. This logic, however, does not apply to

executive options. These options cannot be traded and CEOs cannot hedge (legally) the risk

of their holdings by short-selling company stock. Moreover, CEO compensation contracts typ-

ically grant not only stock options, but also large quantities of company stock to the executive.

As a result, their personal portfolios are likely to be insufficiently diversified. Moreover, their

human capital is invested in their firm, further increasing their exposure to company-specific

17

risk. Thus, a CEO must trade-off the option-value of holding stock options against the costs

of underdiversification. Though the optimal option exercise schedule depends on individual

wealth, diversification, and risk-aversion, a risk-averse CEO should generally exercise options

early given a sufficiently high stock price (Lambert et al., 1991; Hall and Murphy, 2002).

We translate this logic into several measures of overconfidence below. By holding on to exer-

cisable company stock options even when the stock price has exceeded theoretically calibrated

thresholds for exercise, the CEO indicates his belief that the stock price will continue to rise in

the future under his leadership and that this future appreciation will be enough to offset the

benefits of diversifying his holdings. After defining the measures, we explore potential alter-

native explanations for the persistent failure to offset company-specific risk, including inside

information.

Longholder. Our first portfolio measure has two criteria. First, a CEO must at least once

during his tenure as CEO hold an option until the year of expiration. Second, the option held

until expiration must be sufficiently in the money entering its final year to make it unlikely

that the (diminished) option value from holding would exceed the benefits of diversification

achievable through exercise. We use the rational option exercise model of Hall and Murphy

(2002) to guide our choice for this threshold. We use a constant relative risk aversion coefficient

of 3 and 67% of wealth in company stock as parameters in their model to deduce that a CEO

should not continue to hold an option with one year remaining duration if the stock price

is at least 40% higher than the strike price of the option.22 In most cases the option has

already exceeded the threshold for rational exercise for many years (the median percentage

in the money entering the final year for options held to expiration is 253%); however, this

assumption assures that there is at least one point in time at which the CEO’s failure to

exercise is difficult to reconcile with his strong incentive to diversify.23 It also assures that

we do not contaminate our measure by including CEOs who held “underwater” options that

18

could never have been profitably exercised. The particular choice of parameter values is not

important for our results: any assumption from no threshold at all to a threshold of 100% in

the money would yield similar results. Thus, it is CEOs who hold options that are highly in

the money (where option value is relatively small) who drive the results. Holding an option

until its final year, even when it is highly in the money, indicates that the CEO has been

consistently “bullish” about the company’s prospects. Most of the options in our sample have

10 year durations and are fully exercisable after year 4. Thus, for at least 6 years the CEO

decided to hold the options, betting his personal wealth on the company’s future returns,

rather than taking the current value of the option and investing in a diversified portfolio.24

The first version of the Longholder measure captures a managerial fixed effect. We attempt to

capture an inherent personality feature, which makes certain predictions for behavior on the

CEO’s personal and corporate accounts. To test the overconfidence theory, we first show that

this portfolio strategy induces losses and then examine whether the CEOs who engage in the

behavior predicted by overconfidence on their personal account (failure to diversify) are also

the CEOs who engage in the behavior predicted by overconfidence on their corporate account

(excess acquisitiveness). One way to think about the results, then, is as a potential explanation

for the managerial fixed effects on corporate policy identified by Bertrand and Schoar (2003).

If, however, CEOs may become overconfident during their tenures, then the empirical approach

might incorrectly classify earlier years of their tenure among the overconfident CEO years.

Pre-Longholder / Post-Longholder. To address this possibility, we split the Longholder indica-

tor into two separate variables: Post-Longholder is a dummy variable equal to 1 only after the

CEO for the first time holds an option until expiration (provided it exceeds the 40% threshold).

Pre-Longholder is equal to 1 for the rest of the CEO years where Longholder is equal to 1. With

Post-Longholder, we can identify the effect of overconfidence on acquisitiveness using only CEO

years after the CEO has demonstrated overconfidence on his personal account. However, only

19

42% of the observations where Longholder is 1 fall into the Post-Longholder category. Since

there are relatively few completed mergers in the CEO years with Post-Longholder equal to 1

(41), the loss of power is too great to perform tests that require us to subdivide mergers into

finer categories (i.e. cash mergers vs. stock mergers or diversifying mergers vs. intra-industry

mergers).

Holder 67. As an alternative to the Post-Longholder measure, then, we construct Holder 67.25

Here we relax the extreme requirement that CEOs choose to hold their options at every point

in time until expiration. Instead, we consider all option packages with 5 years remaining to

expiration.26 We then ask for each package whether the stock price is sufficiently low (and

hence the option value sufficiently high) to justify continuing to hold the option given risk

aversion and extreme under-diversification. As above, we use the Hall and Murphy framework

as a theoretical guide in choosing a reasonable value for this threshold percentage in the money.

Maintaining the assumed parameter values from the Longholder measure (constant relative risk

aversion coefficient of 3 and 67% of wealth in company stock), the appropriate threshold is 67%

in the money. If the CEO fails to exercise options with 5 years remaining duration when the

potential proceeds from exercise during the preceding year have exceeded the 67% threshold,

we set Holder 67 equal to 1. As above, the results are robust to variation in the value of the

threshold.

We also impose a sample restriction whenever we use this measure. We consider only CEO

years after the CEO for the first time has an option with 5 years remaining duration that is at

least 67% in the money. Thus, we compare CEOs who, faced with highly in-the-money options,

choose to bet on the future stock performance of their company only to (less confident) CEOs

who, faced with the same exercise decision, choose to diversify. Once a CEO enters the sample,

he does not exit unless he leaves the company and once a CEO is classified as overconfident,

he retains that label for the remainder of his sample years. It is possible, however, for a CEO

20

(who has multiple option packages reach the 67% threshold with 5 years remaining duration

during his tenure) to enter the sample in the comparison group, but later become overconfident

under the Holder 67 measure. Overall, the sample restriction leaves 1795 of our original 3911

observations.

Because of the less stringent requirements under the Holder 67 measure, we have more over-

confident CEO years and more completed mergers (154) in those CEO years than under the

Post-Longholder measure. Thus, this measure is more appropriate to test predictions that

require us to partition the mergers into smaller groups. The sample restriction ensures that we

are nevertheless comparing our overconfident CEOs to a less confident set of peers. However,

when we quintile the sample to test financing predictions in Section F, sample size becomes a

limiting issue.

B Discussion

In Panel B of Table 1, we show CEO summary statistics for subsamples of firm years with

and without an overconfident CEO. In Table 2, we report the pairwise correlations between

overconfidence and firm and CEO characteristics. For brevity, we use only Longholder in these

comparisons. The patterns are similar for the other overconfidence measures.

There is little correlation of our overconfidence measures with firm and CEO characteristics.

The only two variables with a correlation higher than 0.1 with Longholder are CEO tenure and

vested options. These correlations arise mechanistically. Since classification as a Longholder

requires the CEO to hold an option (typically for ten years) to expiration, CEOs with short

tenure are less likely to be classified as overconfident.27 This correlation, however, does not

arise with the Holder 67 measure. The (untabulated) correlation between Holder 67 and tenure

is −0.012. Similarly, classification as overconfident under either measure requires the CEO to

21

excessively hold stock options. This failure to exercise naturally can aggregate into a higher

level of option holdings among overconfident CEOs (the correlation between Holder 67 and

vested options is also 0.19). Regardless of the explanation for these correlations, we control

for both tenure and the level of vested options in our estimations.

The correlation between our overconfidence measures and the level of vested options suggests

that we might also be able to measure overconfidence effects simply using the level of stock

and option holdings. Following our logic from the previous section, CEOs who fail to divest

company risk should have higher holdings of stock and options. These potential measures of

overconfidence, however, are less precise and harder to interpret. The CEO cannot fully control

the level of his stock and option holdings due to the influence of the board of directors and his

compensation contract. Thus, our measures of overconfidence focus on changes in the portfolio

that are directly under the CEO’s control. Further, boards grant stock and options to the CEO

primarily to confer incentives. Thus, the effect of vested option levels on acquisitiveness would

confound the overconfidence effect with incentive effects. And, to the extent that mergers are

undesirable, these two effects may go in opposite directions.

We also consider the correlation between the Longholder and Holder 67 measures. Imposing

the Holder 67 sample restriction, we find that the correlation between these two measures is

0.47. Though the Holder 67 measure appears less conservative than Longholder (100 out of

747 of CEOs are classified as overconfident under Longholder; 244 out of 408 under Holder

67), this strong positive correlation suggests that the measures indeed capture the same effect.

Before turning to the effects of our overconfidence measures on investment, we consider alter-

native interpretations of the systematic failure to divest company risk.

1. Inside information. Another important reason a CEO may fail to decrease exposure to

company risk is inside information. CEOs may delay the exercise of vested options beyond

22

the rational benchmark when they have positive inside information about the prospects of

their company. In order to explain repeated delay of option-exercise over a multi-year horizon,

this inside information must be persistently positive over time. Then, one argument against

the information story is that we typically think of the arrival of new information as random

(i.e. sometimes good news and sometimes bad) and it may be implausible for a single piece

of positive information to elude the market for more than six years. The most direct way to

distinguish overconfidence from inside information, however, is to compare the returns CEOs

obtained holding on to their options to the returns they could have obtained through timely

diversification. In Panel A of Table 4, we calculate the hypothetical returns that Longholder

CEOs could have realized had they exercised their options even one year before expiration

and invested the proceeds in the S&P 500. We assume that both the hypothetical exercise

and actual exercise occur at the maximum stock price during the fiscal year. We find that,

on average, Longholder CEOs did not profit by holding until expiration compared to this

alternative strategy. Indeed, the average return to exercising a year earlier is positive, though

statistically insignificant. We also replicate these results assuming hypothetical exercise 2, 3, 4,

and 5 years before expiration.28 The average CEO would have done better under all four

alternative strategies than by holding to expiration. We also make a similar calculation for the

Holder 67 measure. We find, again, no evidence that CEOs have positive information about

future stock prices. In this case, we compute the returns from exercising in year 5 when the

option has passed the 67% threshold and investing the proceeds in the S&P 500. We then

compare those returns to the returns the CEO actually obtains by holding the options until

the next date on which he exercises any options in the package. The mean difference in returns

is −0.0053 with a standard deviation of 0.2882. Thus, there is no evidence that CEOs who

hold in-the-money stock options earn abnormal returns, even over the S&P 500 index. The

CEO’s belief that the firm is undervalued is not correct.

23

Moreover, to be a viable alternative explanation of our results, inside information must explain

not only failure to diversify on the personal portfolio, but also the differences in merger deci-

sions we observe among these CEOs (see Sections III.D-F and IV). To explain heightened

acquisitiveness, positive inside information must be related to upcoming or recently com-

pleted mergers. However, existing empirical evidence suggests no abnormal insider trading

around merger announcements (Boehmer and Netter, 1997). In our empirical work, the Post-

Longholder measure allows us to isolate increased acquisitiveness after the CEO has held an

option to expiration and demonstrated overconfidence on his personal portfolio. Since this

extra acquisitiveness occurs after the options expire, the timing is difficult to reconcile with

the information story.

2. Signalling. Another possible rationale for holding stock options beyond rational bench-

marks for exercise, given informational asymmetry between the CEO and the market about

merger bids, is signalling. Holding exercisable stock options that are highly in the money may

signal to the market that an impending merger bid is high quality. First, existing evidence

on insider trading around merger announcements casts doubt on managers’ ability to signal

merger quality by their trading decisions (Boehmer and Netter, 1997). We also find increased

acquisitiveness among CEOs who excessively hold options not just while they are holding the

options, but also after the options have expired. Most importantly, we show in Section IV that

the market reacts significantly more negatively to the merger bids of CEOs who fail to exercise

highly in-the-money stock options than to the bids of other CEOs. Thus, holding options does

not appear to convey positive information about the merger to the market.

3. Board Pressure. A related story is that CEOs do not exercise highly in-the-money options

because of board pressure. One obvious weakness of this story is the existence of a mechanism

by which the board can compel the CEO not to exercise: they can simply grant options with

longer vesting periods. However, we see almost no cases of vesting periods longer than four

24

years in the data. And, more than half of the option packages are fully exercisable within a

year or less of the grant date. Moreover, to obtain excess acquisitiveness among the CEOs

facing this board pressure, the most natural story would be that the firm plans to do one

or more mergers and fears the market would interpret exercise as a bad signal about those

mergers. But, in order for this pressure to be rational, the market should then prefer the

merger deals of option-holders to exercisers. Again, we see in Section IV that the opposite

is the case. Then, either the board pressure is irrational or the market would punish these

firms even more for merger bids if the CEO did not signal (despite the fact that other firms in

which the CEO exercises options are not punished when they make merger bids). We cannot

completely rule out this convoluted version, but we do the best we can to control for the effects

of board composition and governance on acquisitiveness. We include firm fixed effects in our

regressions so we can identify differences in acquisitiveness between CEOs who hold and do

not hold options within the same firm and, likely, facing a similar board. We also include

controls for board size or, in untabulated results, the number of outside CEOs on the board in

the regressions.

3. Risk Tolerance. Another alternative explanation for our measure is that we have over-

estimated the importance of CEO risk aversion. CEOs might hold options until expiration

if they are risk neutral or even risk-loving, or if they manage to perfectly hedge the risk of

their options despite the prohibition of trading and short sales. However, shareholders should

prefer an (effectively) risk neutral CEO over a risk-averse CEO since they are not prevented

from diversifying their portfolios. So, if risk aversion dampens the willingness of the CEO to

take risks on the corporate account and our overconfidence measures capture risk proneness

or risk neutrality (or a greater ability to offset risk), the market should react positively to the

extra bids of option holders. In Section IV, we show that, instead, the market reacts more

negatively to the bids of CEOs who fail to exercise in-the-money options than to the bids of

25

other acquirors.

4. Taxes. Taxes may provide an alternative reason to postpone option exercise. Specifically,

the CEO may fail to exercise to postpone the payment of taxes on his profits. Personal income

tax deferral, however, would not predict heightened acquisitiveness among holders. Thus, taxes

cannot explain our results.

5. Procrastination. Finally, CEOs might fail to exercise in-the-money options if they are “in-

ertial” in the sense of O’Donoghue and Rabin (2001). We find, however, that more than 68%

of the CEOs classified as overconfident under the Longholder measure (which is most suscep-

tible to the procrastination critique) conduct other transactions on their personal portfolios

in the two years prior to the year their “longheld” option expires. In addition, “inertia” is

less pertinent for Holder 67 since most CEOs who hold beyond the benchmark in the fifth

year nevertheless exercise before the option reaches its final year. Regardless, it is difficult

to explain why holders would be significantly more acquisitive than exercisers, if procrastina-

tion explains late exercise. We might more naturally expect CEOs who procrastinate on their

personal account to also enjoy the “quiet life” on the corporate account.

As a final way to separate the overconfidence interpretation from these other interpretations

of the failure to divest company-specific risk, we construct an alternative measure of over-

confidence. Rather than using the CEO’s decisions on his personal account, we use market

perception of the CEO (as revealed in the business press) to identify overconfidence. The

results using this alternative measure, which side-steps many of these issues, are in Section G.

We also briefly address some additional alternative stories in Section D. We investigate these

stories (e.g. stock price bubbles) using additional controls in our logit specification and, thus,

they are best deferred until after we present the logit results.

26

C Empirical Specification

To test the effect of managerial overconfidence on acquisitiveness, we use the following general

regression specification:

(3) Pr{Yit = 1|Oit, Xit} = G(β1 + β2Oit +X 0itB)

O is the overconfidence measure. The set of controls X includes Tobin’s Q, cash flow, size, a

measure of corporate governance, ownership, unexercised vested options (normalized by total

number of shares outstanding) and year fixed effects. Y is a binary variable that, unless

otherwise specified, takes the value 1 if the CEO made at least one successful merger bid in a

particular firm year. Throughout the paper, we assume that G is the logistic distribution.29

The null hypothesis is that β2, the coefficient on overconfidence, is equal to zero.

There are two kinds of variation we can use to identify the effect of overconfidence on ac-

quisitiveness, cross-sectional and within-company variation. As an example for the first type,

consider the case of Wayne Huizenga, CEO of Blockbuster Entertainment Group for all 7 years

the firm appears in our data. Since he holds some options until the year of expiration, we clas-

sify him as overconfident. He also, during those 7 years, conducts 6 acquisitions. Similarly,

David Farrell is CEO of May Department Stores — the holding company of Lord & Taylor,

Filene’s, and Robinsons-May, among others — for the 15 years it appears in our sample and

is classified as overconfident. He conducts 5 mergers during those 15 years. By contrast, J.

Willard Marriott of Marriott International is CEO of his company for all 15 years of our sam-

ple, but never holds an option until expiration. He also never conducts an acquisition. By

comparing these two types of CEOs, we can identify a cross-sectional effect of overconfidence

on acquisitiveness. As an example of within-company variation, consider Colgate Palmolive.

For the first 4 years, the CEO is Keith Crane. Crane never holds an option until expiration

and he never conducts an acquisition. Reuben Mark succeeds him as CEO in 1984. Over

27

the next 11 years, he holds some options until the year of expiration and he also conducts

4 acquisitions. So, by comparing overconfident and rational CEOs within the same firm, we

might also identify a positive effect of overconfidence on acquisitiveness.

We estimate Equation (3) using three estimation procedures. The first specification, a logit

regression, makes use of both types of variation. The second, a logit regression with random

effects, also makes use of both types of variation. But, it explicitly models the effect of the firm,

rather than the CEO, on acquisitiveness. Note that if the estimated effects of overconfidence

in the logit specification were due to firm effects, we would expect to see a decline in our

estimates when we include random effects. Finally, we estimate Equation (3) using a logit

regression with fixed effects. This specification makes use only of the second type of variation.

That is, we estimate the effect of overconfidence on acquisitiveness using only variation between

overconfident and rational CEOs within a particular firm. To estimate the fixed effects model

consistently, we use conditional logit. Conditioning the likelihood on the number of successes

in each panel, we avoid estimating the coefficients of the fixed effects themselves and obtain

consistent estimates of the remaining coefficients. The fixed effects approach eliminates any

time-invariant firm effect on average acquisitiveness. The disadvantage of the procedure is

that it induces sample-selection bias. Only firms that conduct at least one merger during the

sample period and that had at least one overconfident and one non-overconfident CEO are

included in the fixed-effects estimation. In Table 3, for example, the number of observations

drops from 3911 to 2568 and the number of firms from 394 to 225 when we move from the

logit to the fixed effects logit specification. To show that neither cross-sectional variation nor

sample selection are biasing our results, we present the results of all three specifications.

28

D Overall Impact of Overconfidence

We first estimate Equation (3) on our entire sample of firm years. A positive effect of overcon-

fidence on average is not necessary to confirm the predictions of our overconfidence model (see

Section I). However, such a finding would indicate that overconfidence explains a significant

amount of observed merger activity.

Running a baseline logit with only overconfidence as a regressor (and no controls), we already

find a strong and significant impact of overconfidence on acquisitiveness. Using Longholder as

the overconfidence measure, the odds of an overconfident CEO making an acquisition are 1.65

times the odds of other CEOs. More specifically, the odds of a non-overconfident CEO making

an acquisition are 0.118 while the odds for a Longholder CEO are 0.195. Similarly, when we

split Longholder into the Pre-Longholder and Post-Longholder components, we find an odds

ratio of 1.48 on Post-Longholder. And, using Holder 67 as the overconfidence measure, we find

an odds ratio of 1.78. The Longholder and Holder 67 effects are significant at the 1% level and

the Post-Longholder effect is significant at 10%, where standard errors are clustered by firm.

Table 3 contains the results of estimating Equation (3) using the logit, random effects logit, and

conditional logit specifications and each of the proxies for overconfidence, respectively. Stan-

dard errors in the logit specification are clustered by firm. In the conditional logit specification,

standard errors are not clustered at the firm level. However, in a traditional logit specification

with firm dummies, the errors with firm-level clustering are actually slightly smaller than the

errors from the conditional logit specification.

We include the logarithm of assets at the beginning of the year as a control for firm size,

Tobin’s Q at the beginning of the year as a control for investment opportunities, an indicator

for efficient board size as a measure of corporate governance30, and cash flow as a measure of

internal resources. We also include two controls for the incentive effects of holding company

29

stock and options: the percent of company equity held by the CEO at the beginning of the

year and the number of options exercisable within six months of the beginning of the year,

normalized by total shares outstanding. Finally, we include year effects to control for time

trends in the likelihood of conducting a merger. The most consistent effects across specifications

are for cash flow and Q. We find that firms with lower values of Tobin’s Q are more likely to

conduct mergers, suggesting that acquisitions may be a substitute for profitable investment

opportunities.31 More cash flow, on the other hand, leads to more acquisition activity, as

expected if cash eases financing constraints. Among the other controls, the between and within

firm effects appear to go in opposite directions and most estimated coefficients are insignificant.

Notably, size appears to have a mechanical relationship with acquisitiveness within firm. That

is, the assets of a firm are necessarily larger after a merger. We re-run the regressions without

size to verify that this undesirable effect does not interact with the overconfidence estimates.

The effects of these controls appear to be largely orthogonal to the effect of overconfidence.

CEOs who persistently hold options are still significantly more acquisitive on average, regard-

less of the specification. The effect of overconfidence on acquisitiveness is strong and significant

even when we include firm fixed effects and identify overconfidence only using variation across

CEOs in the same firm.32 We also estimate an untabulated specification that includes industry

fixed effects and the interaction of industry effects with the year effects to the regression.33

Industries are defined as the 48 Fama and French industry groups.34 This specification allows

us to control for the possibility that mergers cluster within industries over time, as argued by

Andrade et al., (2001). There is only a negligible impact on the results. Thus, overconfidence

appears to be an explanation of merger activity that generalizes across merger waves.

Including year and firm effects already address any alternative explanation of the results that

relies on market-wide or firm cross-sectional variation. However, there are possible alternative

stories that rely on time-series variation within firms. Here we consider additional controls

30

that can help rule out relevant time-varying firm effects.

First, we consider the possibility that CEOs exploit stock price bubbles and trade their over-

valued equity for the assets of the target company (Shleifer and Vishny, 2002; Dong et al.,

2002). This story can incorporate the observed failure to exercise options if managers want to

reap the benefits of the bubble or to avoid “popping” it with a negative signal. To address

this possibility, we must check whether the probability of doing a merger moves with the stock

price of a particular firm and whether controlling for this effect ameliorates the estimated co-

efficient of overconfidence. So, we estimate Equation (3) adding five lags of stock returns to

our set of controls. We find that our estimates of the effect of longholder on acquisitiveness

are unaffected (Table 5), though the lag of returns does appear to increase acquisitiveness.

Second, we investigate whether CEOs hold options longer than their peers because their com-

panies’ stocks are more volatile. High volatility of the underlying asset increases option value

and the threshold for exercise. We can link this behavior to increased acquisitiveness if these

CEOs conduct mergers to diversify the corporate account (Amihud and Lev, 1981). Indeed,

we will show in Section E that much of the acquisitiveness of overconfident CEOs is due to

diversifying mergers. We estimate Equation (3) including our usual controls and adding the

volatility of returns over the prior year as an additional control. We find that volatility has no

explanatory power for the time series of merger activity within a firm and our estimate of the

overconfidence effect is virtually unchanged.

Third, we test for effects of dividend policy on the results. CEOs in firms that do not pay

dividends have less incentive to exercise options (option holders do not receive the dividends

from the underlying stock unless they exercise). If firms engaging in mergers and acquisitions

are less likely to pay dividends, then time series variation in dividend payments might explain

our results. To test this story, we re-estimate Equation (3) adding the dividend rate per share

as an additional control. The results are unchanged.

31

Fourth, we test whether other observable personal characteristics might drive both sub-optimal

option exercise decisions and excess acquisitiveness. Specifically, we consider educational back-

ground, age, and CEO tenure. We also consider the effect of awarding the additional titles of

President and Chairman of the Board to the CEO. Tenure is a particularly important control

given the correlation with Longholder reported in Table 2. We find that finance education has

a positive impact on acquisitiveness, but the effect is orthogonal to overconfidence. The other

CEO characteristics (being president and chairman, age, tenure) do not impact the estimated

effect of overconfidence on acquisitiveness (and are not individually significant). Thus, it is

unlikely that our option-holding measures capture an observable CEO characteristic other than

overconfidence.

Finally, as a last test of the information hypothesis, we split the Longholder measure using

the return calculations from Panel A of Table 4. Specifically, we categorize Longholder CEOs

into the group “Did OK” if more often than not when they held an option to expiration they

earned positive profits over the S&P 500. The remaining Longholder CEOs (“should have

exercised”) more often than not would have done better by diversifying their portfolio. We

then re-estimate equation (3) replacing Longholder with these two component variables. We

find that the increased acquisitiveness explained by Longholder is not concentrated among the

CEOs who earn positive profits by holding their options to expiration relative to diversifying

their portfolio (Table 4, Panel B).

Thus, all of the regressions confirm that overconfidence is an important determinant of merger

activity, even on average, and is distinct from other plausible theories of excess acquisitiveness

and option exercise.

32

E Overconfidence and Diversifying Mergers

We have found that overconfident managers, on average, are more likely to make a successful

merger bid than their rational peers. The empirical results suggest that exuberance about

potential merger synergies dominates the countervailing effect of perceived undervaluation,

even on average. We now test the specific predictions of our model of overconfidence.

According to our model, overconfident managers are more likely than rational managers to

undertake a merger project that, ex ante, is unlikely to increase value (Prediction 1). To test

this prediction, we attempt to identify a subset of mergers that, ex ante, is unlikely to create

value. We hypothesize that diversifying mergers are such a subset. Not only is there ample

support in the academic literature for this assumption, but the market also seems to recognize

in advance that many diversifying bids are unwise. Morck et al., (1990) document a negative

market reaction when a firm announces a diversifying deal, an effect we confirm in our data in

Section IV.35

Using diversification as a proxy for mergers with negative expected value, we estimate Equation

(3) with a dependent variable that indicates a successful diversifying bid in a particular firm

year. Bids are defined as diversifying if the acquiror and target firms are not members of the

same Fama-French 48 industry group. We also estimate Equation (3) with a dependent variable

that indicates a successful intra-industry bid. As noted in Section A, the Post-Longholder

measure is not appropriate for tests that require us to partition the mergers into smaller

categories. However, we present the results for both the Longholder and Holder 67 measures.

Table 6 shows that overconfident managers are far more likely to do diversifying mergers than

rational managers, under either measure. In the fixed effects logit specification, the odds ratio

on the Longholder measure of overconfidence is 2.05. The effect of Longholder on the likelihood

of making a related bid is positive (1.44), but insignificant. Similarly, for the Holder 67 measure

33

we obtain an odds ratio of 2.29 for doing a diversifying merger in the fixed effects specification,

but only 1.23 for within-industry mergers.

To gauge the significance of these differences, we re-estimate the regressions using a linear

probability specification within a seemingly unrelated regressions model. We choose this ap-

proach, rather than specifying a multinomial logit, to allow us to estimate the fixed effects

specification. We find that the differences in the effect of overconfidence on the probability of

doing a diversifying or within-industry merger, both with and without fixed effects, are statis-

tically significant at the 10% level for the Longholder measure. With the Holder 67 measure,

the difference without fixed effects is significant at 5%, but the difference with fixed effects is

not statistically significant (the p-value is 0.1776).

Thus, the economically large and statistically significant effect of overconfidence on acquisi-

tiveness is due mainly to overconfident managers conducting more destructive mergers. This

finding confirms Prediction 1 of our model.

F Overconfidence and Internal Resources

Our second prediction is that overconfidence matters most in firms with abundant internal

resources. If a firm can finance an acquisition without issuing equity, perceived undervaluation

by the capital market will have less of an effect on the CEO’s enthusiasm for the merger. Cash

and safe debt allow the CEO and current shareholders to remain the residual claimants on all

of the merger’s future value. Furthermore, an overconfident CEO might prefer risky debt to

equity. While he may disagree with the market about the probability of bankruptcy and, thus,

view debt as too expensive, he retains more rights to the (perceived) upside with risky debt

than with equity. Thus, we predict that the effect of overconfidence on acquisition decisions is

most pronounced in firms with large cash resources and untapped debt capacity.

34

To test this prediction, we employ the Kaplan-Zingales index. Kaplan and Zingales (1997)

use information from annual reports and company executives to measure financing constraints

directly. They then estimate an ordered logit of this classification on five accounting ratios

related to financial constraints. These variables are cash flow to total capital, Q, debt to total

capital, dividends to total capital, and cash holdings to capital. Recent research (Baker et al.

(2001), Lamont et al., (2001), Malmendier and Tate (2003)) uses the estimates to construct

an index of financial constraints (or equity dependence) as follows:

KZit = −1.001909 ∗CFitKit−1

+ 0.2826389 ∗Qit + 3.139193 ∗ Leverageit

−39.3678 ∗Dividendit

Kit−1− 1.314759 ∗

CitKit−1

Higher values of the linear combination of the five ratios implies a higher degree of equity

dependence36. Prediction 2 would be confirmed if the effect of overconfidence is strongest for

the subsample of firms that have the lowest values of the Kaplan-Zingales index.

We divide our sample into quintiles of the Kaplan-Zingales index and estimate random effects

logit regressions of Equation (3) separately on each quintile.37 Since the capital structure

of a firm may change endogenously in anticipation of (or preparation for) a merger, we use

the value of the index at the beginning of the year preceding the merger. The results of our

estimation are in Table 7.38 The dependent variable indicates that the firm made at least one

successful bid in a particular firm year. We find, as predicted, a positive and significant effect

of overconfidence in the “least constrained” quintile (the odds ratio on overconfidence is 2.03)

and no significant effect in the “most constrained” quintile (the odds ratio is 1.07). The large

difference is not due to a lack of sufficient mergers to identify the effect in the most constrained

quintile: the number of successful bids is actually larger in the bottom quintile (72 versus 91).

To test the significance of the difference between the most constrained and least constrained

quintiles, we re-estimate the effects within a single regression using dummy variables for the

35

five quintiles and interacting those dummies with Longholder and the controls (other than the

year effects). The interactions with the control variables allow the coefficients to differ with

financial constraint, as they appear to in Table 9. The difference turns out to not quite be

statistically significant. However, our prediction is on the difference between overconfident and

non-overconfident CEOs given the degree of financial constraint. So, this failure does not affect

our conclusions.

The data confirms Prediction 2 of our model: the effects of overconfidence on acquisitiveness are

strongest for managers with abundant internal resources. The data also confirms the financing

implications of our model. We find that overconfident CEOs are more likely, conditional on

conducting a merger, to finance it using cash and debt (Panel A, Table 8). Here we use both the

Longholder and Holder 67 measures to conduct our tests. The odds ratio of using cash versus

any mixture of risky securities with cash is strongest using Holder 67 (1.38), but is also positive

using Longholder (1.10). We also examine the effect in a regression framework with controls.

In particular, we control for the effects of market over- and undervaluation. We also control for

Tobin’s Q, stock and vested option ownership, merger size, financial constraints and year effects

in various combinations. We find that overconfident CEOs are far more likely than rational

managers to conduct a cash acquisition when the firm is unlikely to be overvalued by the

market, as captured by Tobin’s Q being less than the (within-sample) industry average. Again,

we use the Fama and French 48 industry definitions. The interaction of undervaluation and

overconfidence is strongest with the Longholder measure (and significant across specifications).

With Holder 67, the interaction is always positive, though never significant. However, the

level effect of overconfidence remains reliably positive, despite controlling for the interaction

with market valuation. Interestingly, CEOs do fewer cash deals when they are overvalued

by the market, though the effect is subsumed by financing constraints when we include them

as additional controls in the regression. These result confirm that overconfident managers

36

are particularly sensitive to (perceived) market undervaluation and are also consistent with

the view that investor sentiment affects merger financing decisions, as in Shleifer and Vishny

(2002).

G Overconfidence and the Press

So far, we have used CEOs’ personal portfolio decisions to identify differences in beliefs between

managers and outsiders about the firms’ future prospects. We now assume the perspective of

corporate outsiders — rather than the managerial side — and ask which CEOs the market per-

ceives as “confident” and “optimistic.” Our proxy for market perception uses press coverage in

leading business publications: The Wall Street Journal, The New York Times, Business Week,

Financial Times, and The Economist. Using the press data described in Section II, we record

year by year the number of articles from Factiva.com and LexisNexis searches that refer to the

CEO using the terms (a) “confident” or “confidence,” (b) “optimistic” or “optimism,” (c) “not

confident,” (d) “not optimistic,” and (e) “reliable,” “cautious,” “conservative,” “practical,”

“frugal,” or “steady.” We then compare, for each year, the number of articles that portray

a CEO as confident and optimistic to the number of articles that portray him as not confi-

dent, not optimistic, reliable, cautious, conservative, practical, frugal, or steady. That is, we

construct the following indicator:

TOTALconfident =

⎧⎪⎨⎪⎩ 1 if a+ b > c+ d+ e

0 otherwise

We then relate the TOTALindicator to merger frequency and type of mergers by substituting

the portfolio measures of overconfidence in the previous regressions in two ways. First, we use

the lagged value of TOTALconfindent. Second, we calculate TOTALconfident using all sample

years of a CEO up to (and including) the previous year. In both cases, we control for the total

number of press mentions over the same period and we restrict the analysis to observations up to

37

(and including) the first merger of a CEO. Note that removing coverage frequency is important

both because some CEOs are mentioned more often in the press than others and because a

high number of mentions implies a higher number of mentions as “confident” or “optimistic,”

potentially due to press bias. Using the lagged or cumulative past value of TOTALconfident

and restricting the data set to at most one merger per se deals with the concern that managers

might try to convey confidence and optimism to the press.during a merger. Or, the press may

simply be more likely to perceive a CEO who conducts a mergers as confident and optimistic.

Finally, to make sure that we capture all mergers and press mentions of an executive (in his

role as CEO) we drop executives who became CEO before the beginning of the sample period.

We also attempt to address the concern that personal characteristics other than overconfidence

itself may be driving differential press coverage. As mentioned in Section II, we collected

additional background information for each of the CEOs in the sample. Our hand-collected

information allows us to control for the professional background, age, tenure, and additional

positions of the CEO (chairman, president). Here, finance background is a dummy variable

equal to 1 if the CEO previously worked in a financial institution or as a CFO, treasurer,

accountant or in another finance related position. CEO with engineering background is a

dummy variable equal to 1 if the CEO is an individual patent-holder, or previously worked as

an engineer, in the natural sciences, or in another technically-oriented position. We include

these CEO-level controls on top of the usual firm and ownership controls in each regression.

Table 9 displays the correlations of the press measures and various firm and CEO character-

istics. Panel A reports a positive correlation between TOTALconfident and the Longholder

measure. Both the correlation of the lagged and the correlation of the cumulative version

are statistically significant at the 1% level. The TOTALmention controls display instead in-

significant (zero or negative) correlations. Panels B, C, and D display the correlations of both

versions of the press-based measure with various firm and CEO characteristics.

38

Our press measure of overconfidence not only positively correlates with the longholder measure

of overconfidence, but also performs remarkably similarly in the acquisitiveness regressions.

First, we can replicate the overall acquisitiveness regressions of Table 3, using TOTALconfident

as our proxy for O and the total number of mentions in the press as an additional control (Table

10, Panel A). In the random effects specification, for example, we find odds ratios of 2.8 for

the lagged measure and 2.5 for the cumulative measure, which are significant at the 5% level.

(See columns (5) and (6).) We can also replicate the test of Prediction 1 from Section E,

using diversification as a proxy for negative expected value. Table 10 presents the results.

TOTALconfident, like longholder, predicts a heightened probability of conducting diversifying

deals but not intra-industry mergers. The odds ratios in the random effects specification are

3.4 (lagged measure) and 4.2 (cumulative measure) for diversifying mergers, both significant

at the 5% level, but insignificant for within-industry mergers. Using the same methodology

as in Subsection E, we find that the coefficients of TOTALconfident for diversifying and for

within-industry mergers are significantly different at the 1% level.

We also re-measure the effect of overconfidence conditioning on internal resources (Prediction

2). Estimating Equation (3) separately on quintiles of the Kaplan-Zingales index requires us,

however, to split the already drastically reduced press data set into subsets of about 110 firm-

years. We still find that the odds ratios decrease with increasing financial constraints but fail

to find statistical significance. On the other hand, considering the two highest and the two

lowest KZ quintiles jointly, we replicate the results of Section F.

Finally, we note that press coverage as “optimistic” and “confident” not only predicts acquis-

itiveness, but also strongly predicts increased sensitivity of corporate investment to cash flow,

particularly among the most equity dependent firms. Malmendier and Tate (2003) employ

similar portfolio measures of overconfidence and replicate all results with a simplified version

of the TOTALconfident measure (calculated only once for the full sample period per CEO).

39

These results support the overconfidence story in several ways. First, they show that, whether

we measure differences in beliefs between the manager and the market using managerial portfo-

lio decisions or market perception (as reflected in press coverage), the effect on merger activities

is the same. The additional findings thus strengthen the interpretation of our portfolio-based

measure as overconfidence. Second, our theory assumes that outside financiers are less opti-

mistic about the firm’s future performance and will not provide capital at the rates the CEO

believes are appropriate. Our press results confirm that the market recognizes managerial

overconfidence. Finally, the press results corroborate the view that our overconfidence mea-

sures capture aspects of the CEOs’ personalities rather than an omitted firm effect. While we

have addressed this possibility for the portfolio measures using controls and firm fixed effects,

the press measure provides direct evidence: the searches are for executive personality features.

Framed differently, the press results provide a crucial insight into the type of executive captured

by our portfolio measures of overconfidence.

IV Market Reaction to Overconfidence

Studying mergers and acquisitions provides the opportunity to identify the market’s reaction

to the announcement of the deal. Because many other corporate decisions, like investment,

must be studied in aggregate due to data limitations, we cannot deduce the reaction of the

market to any particular project. With mergers, we know the exact date of announcement.

This allows us to measure market response using daily stock returns.

Our theory predicts that the market will react more negatively to the announced bids of

overconfident CEOs than to the bids of other CEOs (Prediction 3). The negative impact of

overconfidence reflects that overconfident CEOs do some value-destroying mergers and that

they forego some value-creating ones when perceived financing costs are too high. Further,

40

competition can induce overconfident CEOs to overpay for their mergers.

We apply event study methodology (Brown and Warner, 1980 and 1985, and MacKinlay, 1997)

to measure the effect of overconfidence on announcement returns. Here we include all merger

bids, both successful and unsuccessful, in the estimation. The event window is the three days

surrounding the announcement of the bid, starting at day −1 and ending on day +1 where

day 0 is the day of the announcement.39 We calculate the cumulative abnormal return to the

acquiring firm’s stock over this window. Following Fuller et al., (2002), we use market returns as

our proxy for expected returns. This approach is appropriate since our sample consists of large

U.S. companies that compose a substantial portion of market returns. Moreover, we avoid

having to drop overlapping events (as is common in alternative event study methodologies

using estimation periods). In fact, rapid succession of multiple acquisitions may indicate a

particularly high level of overconfidence. Since merging companies is often highly disruptive —

labor forces must be consolidated, corporate cultures must be adapted, etc. — it may be the

height of hubris to juggle several such projects at once.40 So, assuming that α = 0 and β = 1

for the firms in our sample, abnormal returns are given by

ARit = rit − rmt

where rit is firm i’s return on day t of the event window and rmt is the return on the S&P 500

index that day. Cumulative abnormal returns are

CARi =Xt

ARit

To test whether overconfidence has a negative contribution to the mean cumulative abnormal

return during the event window, we run the following cross-sectional regression:

(4) CARi = γ1 + γ2Oi +X 0iG+ εi

where O indicates an overconfident manager and X is the set of controls. The null hypothesis is

γ2 < 0. Table 11 presents the results. We tabulate three main specifications of the regression.

41

First, we compute the baseline difference in CARs for overconfident CEOs relative to other

CEOs, including no controls in the regression. Second, we add controls for the effects of stock

and option ownership (incentives), relatedness of the acquisition (an indicator equal to 1 if

the acquiror and target share the same Fama-French industry group), corporate governance

(efficient board size), and cash financing. Third, we add year effects to control for time trends

in the average market reaction to merger bids.

We also include an additional specification that interacts overconfidence with cash financing

for the Holder 67 and Longholder measures. It is interesting from a governance perspective

to see whether limiting free cash flow might root out the worst merger bids by overconfident

CEOs. There is no clear prediction from the model, however, that overconfident cash mergers

must be the worst deals.

The control variables in the regressions all have the expected signs. The two consistently

significant controls are for cash financing (cash deals — on average — are viewed more favorably

by the market) and vested options. We find that the effect of vested option holdings on

cumulative abnormal returns is decidedly non-linear. We, therefore, include a quadratic term

in vested options as an additional control. The positive incentive effects of vested options

appear maximal in the lower range of vested option holdings. A negative effect — perhaps due

to entrenchment or other negative aspects of excessive CEO power — apply at very high values.

Most importantly, we find strong evidence that the market reacts more negatively to the merger

bids of CEOs who fail to exercise in the money stock options. In the Longholder regressions,

we see a discount of roughly 70 to 100 basis points for overconfident bids, depending on the

specification. Given a baseline negative announcement effect of 40 basis points (see Table 1,

Panel D), the additional discount for mergers of overconfident CEOs is large.

Though we found in the logits of Section D that the effect of Longholder comes from both

42

the Pre-Longholder and Post-Longholder components (with insignificant differences between

the estimated coefficients of the two variables), the same is not true here. Overconfidence

appears to be a fixed effect on managerial acquisitiveness; however, the market only strongly

discounts the bids of overconfident CEOs after they have revealed their overconfidence (at least

partially) through their portfolio decisions. That is, only Post-Longholder has a significant

negative impact on the cumulative abnormal returns to merger bids.

The evidence with the Holder 67 measure is more mixed. We do not find a significant negative

impact of Holder 67 on the cumulative abnormal returns to merger bids on average (though

the effect is negative). However, we do find that the cash deals of overconfident CEOs are

significantly worse, both than the stock deals of overconfident CEOs and than the cash deals

of non-overconfident CEOs. In the Longholder regressions, we also find some evidence that

overconfident cash mergers are the worst deals, though the result is not significant. However,

what is striking is that the total negative impact to an overconfident cash deal is nearly identical

in the Longholder and Holder 67 specifications (roughly 130 basis points). Thus, the difference

in the average impact of overconfidence on cumulative abnormal returns using the two measures

appears to be due to different conclusions about the quality of overconfident stock deals.

Finally, we note that in untabulated regressions we include additional controls for CEO age

and the consolidation of the titles CEO, Chairman of the Board, and President. Both of these

variables appear to negatively impact the cumulative abnormal return to merger bids, but

are orthogonal to the overconfidence effect. We also estimate a specification including the

interaction of industry effects (48 Fama-French industry groups) and year effects as controls.

These additional variables should control for the possibility that the market reacts differently,

on average, to merger bids during a merger wave. But, again, the overconfidence estimates are

unaffected.

43

V Conclusion

The goal of this paper is twofold. First, we show in a simple theoretical framework that the

effect of overconfidence on the baseline frequency of merger activities is ambiguous. Over-

confident CEOs are more eager to make acquisitions, but perceived financing constraints can

prevent them from doing so. However, overconfident CEOs are unambiguously more likely

than rational CEOs to undertake value-destroying acquisitions. And they are more likely to

make acquisitions when their firm has abundant internal resources. Because they do lower

quality deals, on average, and tend to overpay, the market discounts their acquisitions relative

to other CEOs.

We test these predictions using data on a sample of Forbes 500 firms. We classify CEOs as

overconfident if they overinvest in their company with their private funds, and we find that

these CEOs undertake a higher number of takeovers on average, despite the mitigating im-

pact of cash constraints. Further, as predicted by the overconfidence theory, overconfident

CEOs undertake more diversifying mergers, which are unlikely to create value. In addition,

overconfidence has a strong positive impact on the probability of conducting mergers (and par-

ticularly of diversifying mergers) among the least equity dependent firms and no effect among

the most equity dependent firms. These results also hold for CEOs whom the press describes

as “confident” or “optimistic” prior to their merger activities. Finally, the market penalizes

overconfident CEOs for their merger bids: cumulative abnormal returns around overconfident

bids are roughly 100 basis points lower on average than for bids of their non-overconfident

colleagues..

Our results have important implications for contracting practices and organizational design.

Overconfidence provides an alternative explanation for certain agency problems in firms and for

the origin of private benefits. Differently from empire-building preferences, under which CEOs

44

are perpetually and consciously disregarding the interests of the shareholders, overconfident

CEOs believe they are maximizing value. Thus, standard incentives are unlikely to correct

their suboptimal decisions. However, overconfident CEOs do respond to financing constraints.

Overconfidence therefore further motivates the constraining role of capital structure. In ad-

dition, independent directors may need to play a more active role in project assessment and

selection to counterbalance CEO overconfidence.

45

References

[1] Alicke, Mark D; Klotz, M. L; Breitenbecher, David L; Yurak, Tricia J; et al.

“Personal contact, individuation, and the better-than-average effect.” Journal of Person-

ality & Social Psychology, May 1995, 68(5), pp. 804-825.

[2] Alpert, Marc and Raiffa, Howard. “A progress report on the training of probability

assessors.” In: D. Kahneman, P. Slovic, and A. Tversky, Judgement Under Uncertainty:

Heuristics and Biases, Cambridge 1982, pp. 294-305.

[3] Amihud, Yakov and Lev, Baruch. “Risk reduction as a managerial motive for con-

glomerate mergers.” Bell Journal of Economics, 1981, 12, pp. 605-617.

[4] Andrade, Gregor; Mitchell, Mark and Stafford, Erik. “New evidence and perspec-

tives on mergers?” Journal of Economic Perspectives, Spring 2001, 15(2), pp. 103-120.

[5] Asquith, P. “Merger bids, uncertainty, and stockholder returns.” Journal of Financial

Economics, April 1983, 11, pp. 51-83.

[6] Baker, Malcolm; Stein, Jeremy C. and Wurgler, Jeffrey. “When does the market

matter? Stock prices and the investment of equity-dependent firms.” Working Paper,

Harvard University, October 2001.

[7] Baron, Robert A. “Psychological perspectives on entrepreneurship: cognitive and social

factors in entrepreneurs’ success.” Current Directions in Psychological Science, February

2000, 9(1), pp. 15-18.

[8] Bazerman, Max H. Judgment in Managerial Decision Making (5th Edition), Hoboken,

NJ: John Wiley & Sons,Inc., 2002.

[9] Berger, Philip G. and Ofek, Eli. “Diversification’s effect on firm value.” Journal of

Financial Economics, 1995, 37, pp. 39-66.

46

[10] Bernardo, Antonio E. and Welch, Ivo. “On the evolution of overconfidence and

entrepreneurs.” Journal of Economics and Management Strategy, 2001, 10(3), pp. 301-

330.

[11] Bertrand, Marianne and Schoar, Antoinette. “Managing With Style: the Effects of

Managers on Firm Policies.”Quarterly Journal of Economics, 2003, 118(4), pp. 1169-2009.

[12] Black, Fischer and Scholes, Myron. “The pricing of options and corporate liabilities.”

Journal of Political Economy, May-June 1973, 81(3), pp. 637-654.

[13] Boehmer, Ekkehart and Netter, Jeffrey M. “Management optimism and corpo-

rate acquisitions: evidence from insider trading.” Managerial & Decision Economics,

November-December 1997, 18(7-8), pp. 693-708.

[14] Bradley, M.; Desai, A. and Kim, E. H. “The rationale behind interfirm tender offers:

information or synergy?” Journal of Financial Economics, April 1983, 11, pp. 183-206.

[15] Brown, Stephen J. and Warner, Jerold B. “Measuring security price performance.”

Journal of Financial Economics, 1980, 8, pp. 205-258.

[16] Brown, Stephen J. and Warner, Jerold B. “Using daily stock returns: the case of

event studies.” Journal of Financial Economics, 1985, 14, pp. 3-31.

[17] Camerer, Colin and Lovallo, Dan. “Overconfidence and excess entry: an experimental

approach.” American Economic Review, March 1999, 89(1), pp. 306-318.

[18] Camerer, Colin and Malmendier, Ulrike. “Behavioral Economics of Organizations."

Forthcoming in: Economic Institutions and Behavioral Economics, Yrjö Jahnsson Foun-

dation.

[19] Dodd, P. “Merger proposals, managerial discretion and stockholder wealth.” Journal of

Financial Economics, June 1980, 8, pp. 105-138.

47

[20] Dong, Ming; Hirshleifer, David; Richardson, Scott and Teoh, Siew Hong. “Does

investor misvaluation drive the takeover market?” Working Paper, 2002.

[21] Dun & Bradstreet Reference Book of Corporate Managements. Bethlehem PA, Dun &

Bradstreet, Inc. 1997.

[22] Fama, Eugene F. and French, Kenneth R. “Industry costs of equity.” Journal of

Financial Economics, 1997, 43, pp. 153-193.

[23] Fama, Eugene F. and French, Kenneth R. “Testing tradeoff and pecking order

predictions about dividends and debt.” Review of Financial Studies, Spring 2002, 15(1),

p 1-33.

[24] Feather, N. T. and Simon, J. G. “Attribution of responsibility and valence of outcome

in relation to initial confidence and success and failure of self and other.” Journal of

Personality & Social Psychology, May 1971, 18(2), pp. 173-188.

[25] Firth, M. “Takeovers, Shareholder Returns and the Theory of the Firm.” Journal of

Economics, March 1980, pp. 235-260.

[26] Fischhoff, Baruch; Slovic, Paul and Lichtenstein, Sarah. “Knowing with cer-

tainty.” Journal of Experimental Psychology: Human Perception and Performance, 1977,

3, pp. 552-564.

[27] Fuller, Kathleen; Netter, Jeffrey and Stegemoller, Mike. “What do returns to

acquiring firms tell us? Evidence from firms that make many acquisitions.” Journal of

Finance, August 2002, 57(4), pp. 1763-1793.

[28] Goel, Anand Mohan and Thakor, Anjan. “Rationality, overconfidence and leader-

ship.” University of Michigan Business School Faculty Working Paper No. 00-022, 2000.

48

[38] Lambert, Richard A.; Larcker, David F. and Verrecchia, Robert E. “Portfolio

Considerations in Valuing Executive Compensation,” Journal of Accounting Research,

1991, 29(1), pp. 129-149.

[39] Lamont, Owen and Polk, Christopher. “Does diversification destroy value? Evidence

from industry shocks.” Journal of Financial Economics, January 2002, 63(1), pp. 57-77.

[40] Lamont, Owen; Polk, Christopher and Saá-Requejo, Jesus. “Financial constraints

and stock returns.” Review of Financial Studies, 2001, 14, pp. 529-554.

[41] Lang, Larry and Stulz, René. “Tobin’s Q, corporate diversification, and firm perfor-

mance.” Journal of Political Economy, 1994, 102, pp. 1248-1280.

[42] Langer, Ellen J. “The illusion of control.” Journal of Personality and Social Psychology,

1975, 32, pp. 311-328.

[43] Larwood, Laurie and Whittaker, William. “Managerial myopia: self-serving biases

in organizational planning.” Journal of Applied Psychology, 1977, 62, pp. 194-198.

[44] Lys, Thomas and Vincent, Linda. “An analysis of value destruction in AT&T’s ac-

quisition of NCR.” Journal of Financial Economics, 1995, 39, pp. 353-378.

[45] MacKinlay, A. Craig. “Event studies in economics and finance.” Journal of Economic

Literature, 1997, 35, pp. 13-39.

[46] Malmendier, Ulrike and Tate, Geoffrey A. “CEO overconfidence and corporate

investment.” Working Paper, 2003.

[47] March, J. G. and Shapira, Z. “Managerial perspectives on risk and risk taking.”

Management Science, 1987, 33, pp. 1404-18.

[48] Milburn, Michael A. “Sources of bias in the prediction of future events.” Organizational

Behavior and Human Performance, 1978, 21, pp. 17-26.

50

[49] Miller, Dale T. and Ross, Michael. “Self-serving biases in the attribution of causality:

Fact or fiction?” Psychological Bulletin, 1975, 82(2), pp. 213-225.

[50] Moeller, Sara B., Schlingemann, Frederick P. and Stulz, Rene M. "Wealth

destruction on a massive scale? A study of acquiring-firm returns in the recent merger

wave." Journal of Finance, forthcoming.

[51] Moore, P. G. “The manager’s struggle with uncertainty.” Journal of The Royal Statis-

tical Society Series A, 1977, 149, pp. 129-165.

[52] Morck, Randall; Shleifer, Andrei and Vishny, Robert W. “Do managerial objec-

tives drive bad acquisitions?” Journal of Finance, March 1990, 45(1), pp. 31-48.

[53] O’Donoghue, Ted and Rabin, Matthew. "Choice and Procrastination" Quarterly

Journal of Economics, 2001, 116, pp. 121-61.

[54] Ofek, Eli and Yermack, David. “Taking stock: equity-based compensation and the

evolution of managerial ownership.” Journal of Finance, June 2000, 55(3), pp. 1367-1384.

[55] Roll, Richard. “The hubris hypothesis of corporate takeovers.” Journal of Business,

April 1986, 59(2), part 1, pp. 197-216.

[56] Ruback, R. S. and Mikkelson, W. H. “Corporate investments in common stock.”

Proceedings of the Seminar on the Analysis of Security Prices November 1984, 29(2), pp.

179-209.

[57] Schelling, Thomas. The Strategy of Conflict. Cambridge MA: Harvard University Press

1960.

[58] Servaes, Henri. “The value of diversification during the conglommerate merger wave.”

Journal of Finance, 1996, 51, pp. 1201-1225.

51

[59] Shefrin, Hersh. Beyond Greed and Fear : Understanding Behavioral Finance and the

Psychology of Investing. Cambridge: Harvard Business School Press 2000.

[60] Shleifer, Andrei. Clarendon Lectures: Inefficient Markets. Oxford: Oxford University

Press 2000.

[61] Shleifer, Andrei and Vishny, Robert W. “Stock market driven acquisitions.” Work-

ing Paper, June 2002.

[62] Svenson, Ola. “Are we all less risky and more skillful than our fellow drivers?” Acta

Psychologica, 1981, 47, pp. 143-148.

[63] Weber, Roberto A. and Camerer, Colin..“Cultural conflict and merger failure: An

experimental approach.” Management Science, 2003, 49(4), pp. 400-415.

[64] Van den Steen, Eric. “Organizations beliefs and managerial vision.” Mimeo, Stanford

2001.

[65] Weinstein, Neil D. “Unrealistic optimism about future life events.” Journal of Person-

ality and Social Psychology, 1980, 39, pp. 806-820.

[66] Weinstein, Neil D. and Klein, William M. “Resistance of personal risk perceptions to

debiasing interventions.” in Thomas Gilovich, Dale Griffin, and Daniel Kahneman (eds.),

Heuristics and Biases: The Psychology of Intuitive Judgment, Cambridge, UK: Cambridge

University Press, 2002, pp. 313-323.

[67] Weston, J. Fred; Chung, Kwang S. and Sui, Juan A. Takeovers, Restructuring,

and Corporate Governance. Upper Saddle River, NJ: Prentice Hall 1998.

[68] Who’s Who in Finance and Industry. Chicago, Marquis Who’s Who 1980/81-1995/96.

[69] Yermack, David. “Do corporations award CEO stock options effectively?” Journal of

Financial Economics, October 1995, 39(2-3), pp. 237-269.

52

Notes

∗We are indebted to Brian Hall, Kenneth Froot, Mark Mitchell and David Yermack for providing us with

essential parts of the data. We are very grateful to Jeremy Stein and Andrei Shleifer for their invaluable sup-

port and comments. We also would like to thank Gary Chamberlain, David Laibson and various participants in

seminars at Harvard University, Stanford University, University of Chicago, Northwestern University, Wharton,

Duke University, University of Illinois, Emory University, Carnegie Mellon University, INSEAD and Humboldt

University of Berlin for helpful comments. Becky Brunson, Justin Fernandez, Camelia Kuhnen, and Felix Mom-

sen provided excellent research assistance. Author 1 acknowledges support from the Russell Sage Foundation

and the Division of Research of the Harvard Business School. Author 2 acknowledges support from the Russell

Sage Foundation and the Center for Basic Research in the Social Sciences (Harvard University).

1Quote taken from Weston et al., (1998).

2Moeller et al. (forthcoming) find net losses of $134bn at announcement for mergers from 1998 to 2001.

Andrade et al. (2001) find net 1.9% positive announcment net announcement effect for the prior 25 years.

Whether the net effect is significantly positive also varies with the data used (SDC, CRSP) and the event study

methodoly employed. See Jensen and Ruback (1983) and Roll (1986) for surveys of earlier studies.

3See, e.g. Asquith (1983), Bradley et al. (1983), and Andrade et al., (2001) for target gains and Dodd (1980),

Firth (1980), and Ruback and Mikkelson (1984) for acquiror losses. Andrade et al. (2001) find a negative, but

insignificant effect on the acquiror’s value, and Asquith (1983) finds no significant pattern.

4US Newslink December 13, 2001 (“Enron’s Bust: Was it the result of Over-Confidence or a Confidence

Game?”); CFO Magazine June 1, 2004 (“Avoiding decision traps”); Accenture Outlook Journal January 2000

(“Mergers & Acquisitions: Irreconcilable Difference”).

5Hayward and Hambrick (1997) and Hietala et al., (2002) also relate acquisitiveness to CEO hubris. Heaton

(2002) provides a modelling framework for overconfidence and corporate investment.

6See Larwood and Whittaker, 1977; Svenson, 1981; Alicke et al. 1995; Weinstein and Klein, 2002. A different

form of overconfidence is analyzed in the calibration literature; i.e., individuals also tend to overestimate the

accuracy of their beliefs (Fischhoff et al., 1977; Alpert and Raiffa, 1982).

7We follow the literature on self-serving attribution and on the “illusion of control” and assign the labels

53

“overconfidence” to the overestimation of one’s own abilities (such as IQ or driving skill; see Feather and Simon

1971, Langer 1975) and “overoptimism” to the overestimation of exogenous outcomes (such as the outbreak of

a war, see Milburn 1978, Hey 1984, and Bazerman 2002).

8See e.g. Lambert et al., (1991).

9Lang and Stulz (1994), Berger and Ofek (1995), Servaes (1996), and Lamont and Polk (2002), e.g., show

that diversified firms trade at a disount relative to stand-alones in the same line of business.

10Schelling (1960), Goel and Thakor (2000), Bernardo and Welch (2001), and Van den Steen (2001) explore

positive effects of overconfidence.

11See Shleifer (2000).

12For an overview see Camerer and Malmendier (forthcoming).

13Risky debt has similar properties: mangers view the demanded interest rate as too high.

14More generally, the perceived synergies be might depend on the outflow of cash c. In particular, allowing beto decrease with c is a way to capture the dynamic effects of cash constraints (perceived undervaluation) on an

overconfident CEO’s future merger and investment decisions. As long as be(·) > 0, the results of the section gothrough.

15We ignore the knife-edge case of a tie.

16Another potential use of internal resources is to repurchase shares the overconfident CEO perceives to be

undervalued. However, since any gain to remaining shareholders by repurchasing undervalued shares is offset

by a loss to the former shareholders, a CEO who maximizes current shareholder value will not undertake such

a transaction.

17Few of our 477 sample firms are targets; even fewer are acquired by another sample firm.

18This criterion essentially excludes IPOs from our sample. Thus, the more stringent restrictions on insider

trading associated with such firms, such as lockup periods, do not apply.

19All of our results, however, are robust to using only the CRSP merger database, i.e. mergers involving

publicly traded U.S. targets.

20This selection criterion is especially important in our context since we merge data from the SDC database

54

with the CRSP merger data. Acquisitions of small units of another company differ substantially from the

acquisition of large NYSE firms and may not require the direct involvement of the acquiring company’s CEO.

21Definitions of Q and its components as in Fama and French (2002).

22The Longholder measure in Malmendier and Tate (2003) does not add this additional restriction. Adding

the restriction here does not have much impact on the results (see Figure 1, e.g., in the NBER working paper

version of this paper #10807).

23We do not calculate a separate threshold for every option package in our sample — depending on the CEO’s

wealth, diversification, and risk aversion. As we cannot observe each CEO’s degree of risk aversion and wealth

or the fraction of his total wealth invested in company equity, individual calibration would introduce a great

deal of observation-specific noise into the estimation without clear benefits.

24Note as a rough measure of the stakes involved for the CEO, we multiply the current stock price times the

number of options remaining in the package entering the expiration year. The average value is $5,465,086.

25The definition of Holder 67 here has several differences from the definition in Malmendier and Tate (2003),

though the basic intuition is the same. The biggest difference is the removal of all forward-looking information

from the definition, per the suggestion of the referee.

26When we move towards the grant date in defining overconfidence, we need to worry about when the vesting

period ends. The reason we do not go back further than year 5 is that the vast majority of options in our sample

are 10 year options that are fully vested after year 4. If we considered year 4 or earlier, then, we would have

to worry about the exact vesting date and whether heterogeneity in the length of the vesting period (and hence

the point in time at which we could consider the decision to exercise) might affect our conclusions.

27Note that unidentified overconfidence among short-tenured CEOs may attenuate our estimates of the Long-

holder effect on acquisitiveness.

28We also increase the threshold for inclusion in the profits calculation by 0.05 per year to account for the

increase in the Hall-Murphy threshold as remaining duration on the option increases. That is, we only propose

that the CEO exercise if the option is beyond the relevant Hall-Murphy benchmark (approximately) for the

year in question.

29Wherever econometrically possible, we confirmed the robustness of the estimates to the assumption that G

is normal.

55

30The corporate governance literature suggests that an effective board should have no more than 12 members.

The results are robust to the using the logarithm of board size or the number (or percentage) of CEOs of other

companies sitting on the board as alternative measures of governance.

31This effect appears to be non-monotonic. For example, we find a positive and marginally significant coeffi-

cient when we include a dummy variable for “high Tobin’s Q.” (Q > 1) Alternatively, including the square of

Tobin’s Q reverses the direction of the level effect (though it remains insignificant).

32We note, again, that it is possible in the Holder 67 specification for a CEO to begin in the sample as

non-overconfident, but later become overconfident. So, Holder 67 in the fixed effect specification is identified

using both variation within firm across CEOs and also (potential) variation within the CEO.

33Here standard errors are adjusted for clustering within industry, rather than firm.

34See Ken French’s website (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html) for

definitions.

35Further suggestive evidence comes from Lys and Vincent (1995) and Shefrin (2000), who chronicle AT&T’s

1990 acquisition of NCR using exactly this paradigm. Reassuringly, the longholder measure identifies AT&T’s

CEO (Robert Allen) as overconfident.

36For this test, we use the definition of Q employed by Kaplan and Zingales (1997) to avoid rendering the

weights meaningless. The COMPUSTAT data items are: cash flow to capital = (item 18 + item 14) / item 8 ;

Q = [item 6 + (item24 * item 25) - item 60 - item 74] / item 6 ; debt to capital (leverage) = (item 9 + item

34) / (item 9 + item 34 + item 216) ; dividends to capital = item21 + item 19) / item 8 ; cash to capital =

item 1 / item 8. Item 8, capital, is always taken at the beginning of the year (lagged).

37The effects of a simple logit are similar. Fixed effects logit is not feasible since quintiling the sample leaves

us with too few identifiable cases in some subsamples.

38Sample size and power issues make the parallel estimates using Post-Longholder and Holder 67 unreliable;

nevertheless, in each case we observe a higher odds ratio among the most unconstrained quintile than the most

constrained quintile.

39While the three-day window minimizes the effect of any noise in our proxy for expected returns, we find

similar results using a window of five days (−2 to +2).

56

40Nevertheless, the market-model results are almost identical.

57

Obs. Mean Median St. Dev.3,959 5,950.49 2,234.77 13,907.913,959 2,258.94 855.30 5,557.833,742 382.47 151.66 948.643,959 447.46 189.75 963.923,959 0.37 0.26 0.373,959 0.11 0.10 0.073,959 1.42 1.12 0.893,959 0.55 1 0.503,942 0.04 0 0.193,942 0.48 0 0.503,942 0.24 0 0.423,942 0.11 0 0.313,942 0.09 0 0.293,942 0.05 0 0.21

Obs. Mean Median St. Dev. Obs. Mean Median St. Dev.3,958 57.53 58 6.74 676 57.61 58 6.283,913 8.53 6 7.38 653 10.65 9 7.023,959 0.38 0 0.48 676 0.35 0 0.483,386 0.17 0 0.37 605 0.13 0 0.343,959 0.02 0.00 0.07 676 0.02 0.00 0.043,959 0.02 0.00 0.11 676 0.07 0.02 0.252,102 0.25 0 0.43 378 0.18 0 0.382,102 0.19 0 0.39 378 0.11 0 0.31

Stock Ownership

Corporate Governance

Cash Flow

Service Industry

FounderPresident and ChairmanCEO Tenure

Technical IndustryManufacturing Industry

Technical Industry is binary and equal to 1 for firms with primary SIC codes 1000-1799, 8711; Finance Industry is equal to one for firms with primary SIC codes6000-6799; Manufacturing Industry equals 1 for firms with primary SIC codes 2000-3999; Transportation Industry includes all firms with primary SIC codes 4000-4999; Trade Industry are SIC codes 5000-5999; and Service Industry are SIC codes 7000-8710, 8712-8720, 8722-8999. Assets, capital, Q, Stock Ownership, andVested Options are at the beginning of the fiscal year; all other variables are at the end. Finance background is binary and equal to 1 if the CEO previously workedin a financial institution or as a CFO, treasurer, accountant, or in another finance related position. Engineering background is binary and equals 1 if the CEO is anindividual patent-holder or previously worked as an engineer, in the natural sciences, or in another technically-oriented position.

Vested Options Finance BackgroundEngineering Background

Number of firms = 400. Financial variables are reported in $m. Q is the market value of assets over the book value of assets. Cash flow is earnings beforeextraordinary items plus depreciation. Stock ownership is the fraction of company stock owned by the CEO and his immediate family. Vested options are the CEO'sholdings of options that are exercisable within 6 months, as a fraction of common shares outstanding. Vested Options are multiplied by 10 so that the mean isroughly comparable to stock ownership. Corporate governance is a binary variable where 1 signifies that the board of directors has between four and twelvemembers.

Table 1. Summary Statistics

Panel B. Summary Statistics of CEO Data

Investment (CAPX)

Panel A. Summary Statistics of Firm Data

Cash Flow normalized by lagged capital (CF/k)Cash Flow normalized by lagged assets (CF/a)

Age

Assets Capital (PPE)

Full Sample (747 CEOs) Overconfident CEOs (100 CEOs)

Q

Transportation IndustryTrade IndustryFinancial Industry

Obs. Mean Median St. Dev.3,936 91.81 39 175.773,936 0.81 0 1.923,936 0.71 0 1.703,936 0.02 0 0.193,936 0.06 0 0.293,914 0.52 0 1.34

Obs. Mean Median St. Dev.865 0.398 0 0.490865 -0.004 -0.007 0.048865 0.023 0 0.150865 0.302 0 0.459865 0.105 0 0.307865 0.073 0 0.260865 0.453 0 0.498865 0.044 0 0.205Acquiror in Service Industry

Acquiror in Transportation IndustryAcquiror in Trade Industry

"Not Optimistic" Mentions"Reliable, Cautious, Conservative, Practical, Steady, Frugal" Mentions

Acquiror in Financial Industry

Acquiror in Technical IndustryAcquiror in Manufacturing Industry

Table 1. Summary Statistics (continued)

Number of firms = 400. Press data comes from Business Week, The New York Times, Financial Times, The Economist and The Wall Street Journal usingLexisNexis and Factiva.com. Relatedness is a dummy variable which takes the value 1 when the acquiror and target share the same Fama-French 48 industrygroup. Cumulative abnormal returns to the acquiror are calculated for an event window of -1 to +1 using a modified market model with the daily S&P 500return as proxy for expected returns. The sample consists of 865 merger bids.

"Optimistic" Mentions

Panel C. Summary Statistics for Press Data (327 firms; 661 CEOs)Total Mentions"Confident" Mentions

RelatednessCumulative abnormal return to

Panel D. Summary Statistics of Mer

"Not Confident" Mentions

Longholder Size Q Cash FlowStock Ownership

Vested Options

Corporate Governance

Longholder 1.00Size -0.09 1.00Q 0.08 -0.32 1.00Cash Flow 0.09 -0.14 0.40 1.00Stock Ownership -0.04 -0.19 0.12 0.13 1.00Vested Options 0.19 -0.16 0.08 0.17 0.09 1.00Corporate Governance 0.03 -0.38 0.13 0.07 0.20 0.08 1.00

Longholder AgePres & Chm Tenure

Longholder 1.00Age 0.01 1.00President and Chairman -0.03 -0.03 1.00Tenure 0.13 0.39 0.01 1.00

Longholder Fin. Ed. Tech. Ed.Longholder 1.00Financial Background -0.07 1.00Engineering Background -0.09 -0.23 1.00

Corporate governance is a binary variable equal to 1 if the board has between 4 and 12 directors. Finance background (dummy) equals 1 if the CEOpreviously worked as CFO, treasurer, accountant, in a financial institution or another finance related position. Engineering background (dummy)equals 1 if the CEO is an individual patent-holder, previously worked as engineer, in the natural sciences or another technically-oriented position.

Panel C. Correlations with CEO Characteristics (II): Educational Background (N =2102)

Table 2. Correlations with Overconfidence Measure

Longholder is a binary variable and equals 1 if the CEO ever held an option package until the year before expiration, unless it was < 40% in the moneyentering its last year. Size is the log of assets, Q the market value of assets over the book value of assets. Cash flow is earnings before extraordinaryitems plus depreciation, normalized by beginning-of-the-year capital. Stock Ownership is the fraction of company stock owned by the CEO and hisimmediate family. Vested Options are the CEO's holdings of options that are exercisable within 6 months of the beginning of the year, as a fraction ofcommon shares outstanding and multiplied by 10 so that the mean is comparable to Stock Ownership. Size, Q, Stock Ownership, and Vested Optionsare measured at the beginning of the year.

Panel A. Correlations with Firm Characteristics (N =3959)

Panel B. Correlations with CEO Characteristics (I) (N =3912)

logitRandom

Effects logitFixed Effects

logit logitRandom

Effects logitFixed Effects

logit logitRandom

Effects logitFixed Effects

logit(1) (2) (3) (4) (5) (6) (7) (8) (9)

Size 0.946 0.9358 0.6537 0.9428 0.9334 0.66 0.9932 0.9844 0.4942(0.95) (1.03) (2.68)*** (1.01) (1.06) (2.61)*** (0.08) (0.18) (2.32)**

Qt-1 0.6476 0.6225 0.7135 0.6465 0.6222 0.7154 0.6664 0.6547 0.7701(4.23)*** (3.99)*** (1.88)* (4.26)*** (3.99)*** (1.86)* (2.93)*** (2.74)*** (1.10)

Cash Flow 1.9143 2.1949 2.0231 1.9196 2.2002 2.0377 1.6238 1.787 1.7283(4.34)*** (4.76)*** (2.08)** (4.36)*** (4.78)*** (2.10)** (2.37)** (2.53)** (1.08)

Stock Ownership 1.4913 1.1862 0.384 1.4593 1.1626 0.3813 0.3297 0.3906 0.0327(0.50) (0.18) (0.70) (0.47) (0.16) (0.70) (0.70) (0.56) (0.81)

Vested Options 1.5125 1.0626 0.4566 1.4798 1.0413 0.4595 3.3821 2.5566 1.093(2.42)** (0.15) (1.64) (2.18)** (0.10) (1.62) (1.48) (1.00) (0.07)

Corporate Governance 0.7569 0.8105 1.0817 0.7592 0.8123 1.0811 1.1563 1.2306 2.1376(2.05)** (1.53) (0.42) (2.03)** (1.52) (0.42) (0.75) (1.08) (2.52)**

Longholder 1.658 1.8292 2.1891(3.15)*** (3.62)*** (2.30)**

Post-Longholder 1.4444 1.538 1.8642(1.76)* (1.89)* (1.54)

Pre-Longholder 1.8259 2.0581 2.3305(3.08)*** (3.71)*** (2.41)**

Holder 67 1.7578 2.0552 2.6558(3.15)*** (3.76)*** (2.83)***

Firm Fixed Effects no no yes no no yes no no yesYear Fixed Effects yes yes yes yes yes yes yes yes yesObservations 3911 3911 2568 3911 3911 2568 1795 1795 986Number of Firms 394 225 394 225 315 140

Longholder is a binary variable where 1 signifies that the CEO at some point during his tenure held an option package until the last year before expiration, provided that the package was at least 40% inthe money entering its last year. Post-Longholder is a dummy equal to 1 for all CEO-years after the CEO for the first time holds options to expiration. Pre-Longholder are all years classified as 1 underLongholder, but 0 under Post-Longholder. Holder 67 is a dummy equal to 1 for all CEO years after the CEO for the first time fails to exercise a 67% in the money option with 5 years remainingduration. In the Holder 67 regressions, the sample is limited to CEO years after the CEO for the first time had a 67% in the money option with 5 years remaining duration. The fixed effects logit modelis estimated consistently using a conditional logit specification. Standard errors in columns 1, 4, and 7 are robust to heteroskedasticity and arbitrary within-firm serial correlation. Coefficients arepresented as odds ratios.

* significant at 10%; ** significant at 5%; *** significant at 1%

Table 3. Do Overconfident CEOs Complete More Mergers?

Robust z statistics in parentheses. Constant included.

The dependent variable is binary where 1 signifies that the firm made at least one merger bid that was eventually successful in a particular firm year. Size is the log of assets at the beginning of the year.Q is the market value of assets over the book value of assets. Cash flow is earnings before extraordinary items plus depreciation and is normalized by capital at the beginning of the year. Stockownership is the fraction of company stock owned by the CEO and his immediate family at the beginning of the year. Vested options are the CEO's holdings of options that are exercisable within 6months of the beginning of the year, as a fraction of common shares outstanding. Vested options are multiplied by 10 so that the mean is roughly comparable to stock ownership. Corporate governanceis a binary variable where 1 signifies that the board of directors has between four and twelve members.

Panel A. Returns

Percentile10th20th30th40th50th60th70th80th90thMean

Standard Deviation

Table 4. Are Overconfident CEOs Right to Hold their Options?

0.390.03

-0.05

0.030.100.19

0.27

For each option that is held until expiration and that is at least 40% in the money at the beginning of its final year, we calculate the return the CEOwould have gotten from instead exercising the option a year sooner and investing in the S&P 500. We assume exercise both in the final year and inthe hypothetical year occur at the maximum stock price during that year.

-0.03

Return-0.24-0.15-0.10

logit

Random Effects logit

Fixed Effects logit logit

Random Effects logit

Fixed Effects logit logit

Random Effects logit

Fixed Effects logit

(1) (2) (3) (4) (5) (6) (7) (8) (9)Size 0.9708 0.9647 0.6481 0.968 0.9626 0.6531 0.9911 0.9862 0.507

(0.50) (0.55) (2.56)** (0.54) (0.58) (2.51)** (0.11) (0.16) (2.16)**

Qt-1 0.5885 0.5544 0.6074 0.5876 0.5542 0.6094 0.5651 0.5463 0.6135(4.40)*** (4.31)*** (2.20)** (4.40)*** (4.31)*** (2.18)** (3.43)*** (3.48)*** (1.63)

Cash Flow 1.7048 1.9283 1.5329 1.7093 1.9337 1.5431 1.3937 1.5382 1.7037(3.15)*** (3.72)*** (1.19) (3.16)*** (3.73)*** (1.20) (1.44) (1.81)* (1.01)

Stock Ownership 0.6843 0.6223 0.2552 0.6707 0.6105 0.2531 0.2027 0.2977 0.0765(0.42) (0.44) (0.79) (0.45) (0.46) (0.80) (0.85) (0.69) (0.67)

Vested Options 3.7237 2.6886 0.853 3.6868 2.6485 0.843 1.865 1.5508 1.2135(1.65)* (1.20) (0.14) (1.65)* (1.18) (0.15) (0.73) (0.46) (0.15)

Corporate Governance 0.745 0.7952 1.129 0.7466 0.7967 1.1289 1.1876 1.2486 2.2015(2.09)** (1.64) (0.64) (2.08)** (1.63) (0.64) (0.88) (1.15) (2.57)**

Returnst-1 1.8244 1.8364 1.5734 1.8194 1.8292 1.5675 1.88 1.8759 1.4211(2.95)*** (2.93)*** (1.94)* (2.93)*** (2.91)*** (1.93)* (2.09)** (2.02)** (0.94)

Returnst-2 1.3241 1.3238 1.1289 1.3233 1.3228 1.1273 1.8242 1.8335 1.4268(1.60) (1.46) (0.57) (1.59) (1.46) (0.56) (2.23)** (2.13)** (1.07)

Returnst-3 1.1158 1.1207 1.0679 1.1166 1.1214 1.0675 1.525 1.6086 1.5792(0.57) (0.60) (0.33) (0.57) (0.61) (0.33) (1.47) (1.73)* (1.41)

Returnst-4 1.3781 1.4092 1.4469 1.3769 1.407 1.4444 1.1727 1.1993 1.3252(1.45) (1.76)* (1.76)* (1.44) (1.75)* (1.76)* (0.46) (0.65) (0.87)

Returnst-5 1.1987 1.1756 1.1215 1.1997 1.1756 1.1209 1.7428 1.6508 1.4586(0.99) (0.84) (0.58) (1.00) (0.84) (0.57) (2.22)** (1.80)* (1.23)

Longholder 1.5722 1.7061 2.1272(2.71)*** (3.18)*** (2.19)**

Post-Longholder 1.4219 1.5074 1.9328(1.67)* (1.80)* (1.60)

Pre-Longholder 1.6899 1.8611 2.2084(2.59)*** (3.14)*** (2.23)**

Holder 67 1.9475 2.2638 2.8113(3.77)*** (4.21)*** (2.98)***

Year Fixed Effects yes yes yes yes yes yes yes yes yesObservations 3681 3681 2439 3681 3681 2439 1758 1793 960Number of Firms 365 211 365 211 307 137

Longholder is a binary variable where 1 signifies that the CEO at some point during his tenure held an option package until the last year before expiration, provided thatthe package was at least 40% in the money entering its last year. Post-Longholder is a dummy equal to 1 for all CEO-years after the CEO for the first time holds optionsto expiration. Pre-Longholder are all years classified as 1 under Longholder, but 0 under Post-Longholder. Holder 67 is a dummy equal to 1 for all CEO years after theCEO for the first time fails to exercise a 67% in the money option with 5 years remaining duration. In the Holder 67 regressions, the sample is limited to CEO yearsafter the CEO for the first time had a 67% in the money option with 5 years remaining duration. The fixed effects logit model is estimated consistently using aconditional logit specification. Standard errors in column 1 are robust to heteroskedasticity and arbitrary within-firm serial correlation. Coefficients are presented asodds ratios.

* significant at 10%; ** significant at 5%; *** significant at 1%

Table 5. Control for Returns

The dependent variable is binary where 1 signifies that the firm made at least one merger bid that was eventually successful in a particular firm year. Size is the log ofassets at the beginning of the year. Q is the market value of assets over the book value of assets. Cash flow is earnings before extraordinary items plus depreciation andis normalized by capital at the beginning of the year. Stock ownership is the fraction of company stock owned by the CEO and his immediate family at the beginning ofthe year. Vested options are the CEO's holdings of options that are exercisable within 6 months of the beginning of the year, as a fraction of common sharesoutstanding. Vested options are multiplied by 10 so that the mean is roughly comparable to stock ownership. Corporate governance is a binary variable where 1signifies that the board of directors has between four and twelve members. Returns are the natural logarithm of 1 plus the annual return on company equity.

Robust z statistics in parentheses. Constant included.

logitRandom

Effects logitFixed

Effects logit logitRandom

Effects logitFixed

Effects logit(1) (2) (3) (4) (5) (6)

Size 1.0849 1.0968 0.7585 0.7772 0.7446 0.4456(1.31) (1.18) (1.37) (3.03)*** (3.14)*** (3.45)***

Qt-1 0.68 0.7014 0.9606 0.6496 0.5878 0.5001(2.77)*** (2.45)** (0.20) (3.55)*** (3.23)*** (2.49)**

Cash Flow 1.9102 2.1723 1.8172 1.8388 2.0455 2.3926(3.90)*** (3.99)*** (1.60) (3.07)*** (3.13)*** (1.62)

Stock Ownership 2.6364 1.5318 0.0692 1.4355 2.0722 2.2829(1.06) (0.36) (1.28) (0.26) (0.64) (0.55)

Vested Options 1.6215 1.4471 0.9125 0.827 0.5201 0.1903(3.00)*** (0.84) (0.17) (0.81) (1.08) (2.10)**

Corporate Governance 0.69 0.7253 0.8347 0.834 0.8675 1.1175(2.21)** (1.91)* (0.79) (0.97) (0.74) (0.43)

Longholder 1.6942 1.8465 2.005 1.3167 1.4166 1.4432(2.89)*** (3.07)*** (1.83)* (1.23) (1.46) (0.73)

Year Fixed Effects yes yes yes yes yes yesObservations 3911 3911 1900 3911 3911 1520Number of Firms 394 167 394 131

Size 1.1387 1.1673 0.4742 0.8244 0.7951 0.442(1.46) (1.43) (1.90)* (1.68)* (1.88)* (1.92)*

Qt-1 0.6191 0.6431 0.8883 0.7117 0.6795 0.6058(2.36)** (2.18)** (0.36) (2.20)** (1.93)* (1.26)

Cash Flow 1.6166 1.7717 0.8871 1.4547 1.5346 2.7408(2.11)** (2.12)** (0.21) (1.42) (1.35) (1.30)

Stock Ownership 2.7453 3.2534 0.0663 0.0221 0.0137 0(0.66) (0.69) (0.62) (1.87)* (1.41) (1.24)

Vested Options 3.5317 3.88 2.1107 2.6078 1.5671 0.1204(0.98) (1.14) (0.47) (1.31) (0.36) (1.07)

Corporate Governance 1.1145 1.1588 1.5289 1.0639 1.0733 2.2568(0.47) (0.63) (1.11) (0.25) (0.28) (1.96)**

Holder 67 1.7166 1.8729 2.2903 1.1504 1.2528 1.232(2.51)** (2.73)*** (1.95)* (0.60) (0.92) (0.45)

Year Fixed Effects yes yes yes yes yes yesObservations 1795 1795 693 1795 1795 581Number of Firms 315 102 315 80

The dependent variable in panel 1 is binary where 1 signifies that the firm made a diversifying merger bid that was eventually successful in a particularfirm year. The dependent variable in panel 2 is binary where 1 signifies that the firm made a within-industry merger bid that was eventually successfulin a particular firm year. Industries are the 48 Fama and French industry groups (1997). Size is the log of assets at the beginning of the year. Q is themarket value of assets over the book value of assets. Cash flow is earnings before extraordinary items plus depreciation and is normalized by capital atthe beginning of the year. Stock ownership is the fraction of company stock owned by the CEO and his immediate family at the beginning of the year.Vested options are the CEO's holdings of options that are exercisable within 6 months of the beginning of the year, as a fraction of common sharesoutstanding. Vested options are multiplied by 10 so that the mean is roughly comparable to stock ownership. Corporate governance is a binary variable where 1 signifies that the board of directors has between four and twelve members. Longholder is a binaryvariable where 1 signifies that the CEO at some point during his tenure held an option package until the last year before expiration, provided that thepackage was at least 40% in the money entering its last year. Holder 67 is a dummy equal to 1 for all CEO years after the CEO for the first time fails toexercise a 67% in the money option with 5 years remaining duration. In the Holder 67 regressions, the sample is limited to CEO years after the CEO forthe first time had a 67% in the money option with 5 years remaining duration. The fixed effects logit model is estimated consistently using a conditionallogit specification. Standard errors in columns 1 and 4 are robust to heteroskedasticity and arbitrary within-firm serial correlation. Coefficients arepresented as odds ratios.

* significant at 10%; ** significant at 5%; *** significant at 1%

Table 6. Diversifying and Same-Industry MergersPanel 1. Diversifying Mergers Panel 2. Within Industry Mergers

Robust z statistics in parentheses. Constant included.

A. Longholder

B. Holder 67

Least Equity Dependent

Most Equity Dependent

Random Effects logit

Random Effects logit

Random Effects logit

Random Effects logit

Random Effects logit

Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5Size 0.8721 1.1371 0.9434 0.7672 0.9828

(1.05) (1.00) (0.39) (1.43) (0.10)Qt-1 0.4601 0.8058 0.62 0.6522 0.6072

(3.05)*** (1.00) (1.46) (1.28) (1.31)Cash Flow 0.9115 1.6897 8.0689 3.973 5.6369

(0.27) (0.98) (2.61)*** (1.38) (2.35)**Stock Ownership 0.1046 0.2199 4.9239 3.1767 2.978

(0.82) (0.59) (0.71) (0.38) (0.74)Vested Options 1.0536 111.1586 2.0983 0.7263 12.3633

(0.06) (2.25)** (0.70) (0.11) (1.62)Corporate Governance 0.924 0.8296 0.447 0.8871 0.9991

(0.24) (0.63) (2.42)** (0.36) (0.00)Longholder 2.0289 1.6269 1.6465 2.1899 1.0654

(2.09)** (1.54) (1.40) (1.58) (0.15)Year Fixed Effects yes yes yes yes yesObservations 731 733 708 735 665Number of Firms 121 157 172 167 149

Vested options are the CEO's holdings of options that are exercisable within 6 months of the beginning of the year, as a fractionof common shares outstanding. Vested options are multiplied by 10 so that the mean is roughly comparable to stock ownership.Corporate governance is a binary variable where 1 signifies that the board of directors has between four and twelve members.Longholder is a binary variable where 1 signifies that the CEO at some point during his tenure held an option package until thelast year before expiration, provided that the package was at least 40% in the money entering its last year. Post-Longholder is adummy equal to 1 for all CEO-years after the CEO for the first time holds options to expiration. All regressions are logit withrandom effects. Coefficients are presented as odds ratios.

* significant at 10%; ** significant at 5%; *** significant at 1%

Table 7. Overconfidence and Acquisitiveness by Equity Dependence

--------------------------------->

z statistics in parentheses. Constant included.

The dependent variable in panel 1 is binary where 1 signifies that the firm made at least one merger bid that was eventuallysuccessful in a particular firm year. The dependent variable in panel 2 is binary where 1 signifies that the firm made at leastone diversifying merger bid that was eventually successful in a particular firm year. Industries are the 48 Fama and Frenchindustry groups (1997). Size is the log of assets at the beginning of the year. Q is the market value of assets over the bookvalue of assets. Cash flow is earnings before extraordinary items plus depreciation and is normalized by capital at thebeginning of the year. Stock ownership is the fraction of company stock owned by the CEO and his immediate family at thebeginning of the year.

Panel A. Longholder

Stock AND

Cash

Debt OR Cash and

Debt

Cash and/or Debt Stock

odds (cash v. stock)

odds (cash v. other)

odds ratio (v stock)

odds ratio (v other)

38.8% 6.9% 19.7% 34.6% 1.12 0.63 1.09 1.1040.0% 7.6% 18.8% 33.6% 1.19 0.67 1.47 1.3833.5% 8.3% 25.6% 32.6% 1.03 0.5829.6% 9.1% 24.7% 36.6% 0.81 0.48

logit logit logit logit logit logit logit logit(1) (2) (3) (4) (5) (7) (8) (9)

Undervalued (UV) 1.7215 1.8244 1.8457 1.9274 0.9073 1.7894 1.7659 1.8733(2.72)*** (2.97)*** (3.02)*** (3.11)*** (0.38) (1.66)* (1.59) (1.53)

Qt-1 1.2532 1.2618 1.0288 0.4706 0.9784 0.9603 0.8908(1.24) (1.26) (0.14) (3.30)*** (0.11) (0.19) (0.49)

Stock Ownership 1.7263 1.6837 0.2223 144.5774 25.8462(0.41) (0.42) (1.10) (1.88)* (1.19)

Vested Options 0.5818 0.4279 0.1464 0.7802 1.0133(0.75) (1.21) (0.83) (0.17) (0.01)

Merger Size 0.981 0.9927 0.9922 1.0391 1.0713(1.37) (0.58) (0.76) (1.30) (2.29)**

KZ Quintile 2 0.7824(0.73)

KZ Quintile 3 0.6403(1.22)

KZ Quintile 4 0.5282(1.70)*

KZ Quintile 5 0.4041(2.57)**

Longholder 0.7423 0.708 0.7685 0.7766 0.6792(0.84) (0.99) (0.76) (0.72) (0.97)

UV * Longholder 2.3096 2.41 2.2577 1.9555 3.1857(2.09)** (2.20)** (2.06)** (1.71)* (2.61)***

Holder 67 1.2928 1.3554 1.3902(0.70) (0.81) (0.84)

UV * Holder 67 1.1543 1.1521 1.1071(0.31) (0.30) (0.20)

Year Fixed Effects no no no yes yes yes yes yesObservations 772 772 772 772 430 405 405 405Robust z statistics in parentheses. Constant included.* significant at 10%; ** significant at 5%; *** significant at 1%

Merger size is the amount the acquiror paid for the target as a fraction of acquiror value (for SDC mergers, amount paid is the value of the transaction; for CRSPmergers, it is the market value of the target the day after the announcement. When both variables are present, we use the minimum). KZ Quintile 'x' is a dummyvariable equal to 1 if the lagged value of the Kaplan-Zingales index for that firm year is in the `x'th quintile. Longholder is a binary variable where 1 signifies that theCEO at some point during his tenure held an option package until the last year before expiration, provided that the package wa-283.8m)he least 40% in the money entering itslast year. Holder 67 is a dummy equal to 1 for all CEO years after the CEO for the first time fails to exercise a 67% in the money option with 5 years remainingduration. In the Holder 67 regressions, the sample is limited to CEO years after the CEO for the first time had a 67% in the money option with 5 years remainingduration. UV * Longholder and UV * Beyond Threshold are interactions. Standard errors are robust to heteroskedasticity and arbitrary within-firm serial correlation.Coefficients are presented as odds ratios.

(0.03)0.9266

(0.03)1.0876(2.76)***1.3553

Panel A. All Mergers with Disclosed Method of Payment

Table 8. Merger Financing and Overconfidence

Panel B. Regressions

Overconfident CEOs

Non-overconfident CEOs

Sample includes all merger bids that were eventually successful. The dependent variable is binary where 1 signifies that the bid was financed using only cash.Undervalued is a binary variable where 1 indicates that Q at the beginning of the year was less than or equal to industry Q, where industries are the 48 Fama-Frenchindustry groups. Q is the market value of assets over the book value of assets. Stock ownership is the fraction of company stock owned by the CEO and his immediatefamily at the beginning of the year. Vested options are the CEO's holdings of options that are exercisable within 6 months of the beginning of the year, as a fraction ofcommon shares outstanding. Vested options are multiplied by 10 so that the mean is roughly comparable to stock ownership.

Overconfidence Measure

LongholderHolder 67LongholderHolder 67

logit(6)

1.8007(1.68)*

1.2895(0.70)

1.1564(0.31)

no405

logit(10)

0.9334(0.13)

0.3876(2.79)***0.911

(0.58)0.7711

(0.49)1.0217

(0.22)

(0.03)0.4097

218

(1.53)

1.1994(0.28)

yes

0.8901

LongholderTOTconf. (lagged)

TOTmen. (lagged) Longholder

TOTconf. (cum.) TOTmen. (cum.)

Longholder 1.00 1.00TOTALconfident (lag.) 0.06 1.00 0.13 1.00TOTALmentions (lag.) 0.00 0.35 1.00 -0.01 0.35 1.00

TOTAL-conf.

(lagged)

TOTAL-ment.

(lagged) Size QCash Flow

CEO Owner-ship

CEO Vested Options

Corporate Governance

TOTALconfident (lag.) 1.00TOTALmentions (lag.) 0.35 1.00

Size 0.17 0.31 1.00Q 0.04 0.04 -0.32 1.00

Cash Flow 0.01 0.04 -0.14 0.40 1.00CEO Ownership 0.00 0.14 -0.19 0.12 0.13 1.00

CEO Vested Options 0.00 -0.01 -0.16 0.08 0.17 0.09 1.00Corporate Governance -0.08 -0.11 -0.38 0.13 0.07 0.20 0.08 1.00

TOTAL-confident

(cum.)

TOTAL-mentions

(cum) Size QCash Flow

CEO Owner-ship

CEO Vested Options

Corporate Governance

TOTALconfident (cum.) 1.00TOTALmentions (cum.) 0.33 1.00

Size 0.22 0.30 1.00Q 0.07 0.04 -0.32 1.00

Cash Flow 0.03 0.04 -0.14 0.40 1.00CEO Ownership 0.03 0.09 -0.19 0.12 0.13 1.00

CEO Vested Options 0.02 0.01 -0.16 0.08 0.17 0.09 1.00Corporate Governance -0.08 0.07 -0.38 0.13 0.07 0.20 0.08 1.00

Q is the market value of assets over the book value of assets at the beginning of the year. Cash flow is earnings before extraordinary items plus depreciation and isnormalized by capital at the beginning of the year. CEO ownership is the fraction of company stock owned by the CEO and his immediate family at the beginning ofthe year. CEO vested options are the CEO's holdings of options that are exercisable within 6 months of the beginning of the year, as a fraction of common sharesoutstanding. Vested options are multiplied by 10 so that the mean is roughly comparable to stock ownership. Corporate governance is the number of directors whocurrently serve as CEOs of other companies.

Panel A. Press Measures with Longholder (N = 3372)

TOTALconfident is a dummy variable equal to 1 when the number of "confident" and "optimistic" mentions for a CEO in the LexisNexis and Wall Street Journalsearches exceeds the number of "not confident", "not optimistic", and "reliable, cautious, practical, conservative, steady, frugal" mentions. TOTALmentions is thetotal number of articles mentioning the CEO in both sets of searches. The "lagged" version considers all articles in the previous year. The "cumulative" versionconsiders all articles over the sample period up to the previous year. Longholder is a binary variable where 1 signifies that the CEO at some point during his tenureheld an option package until the last year before expiration, provided that the package was at least 40% in the money entering its last year. Size is the naturallogarithm of assets at the beginning of the year.

Table 9. Correlations of Press Measure

Panel B. Correlations of Press Coverage with Firm Characteristics (N = 3372)

LongholderTOTconf. (cum.)TOTmen. (cum.)

TOTconf. (lagged)

TOTmen. (lagged) Age

Pres & Chm Tenure

TOTconf. (cum.)

TOTmen. (cum.) Age

Pres & Chm Tenure

TOTALcon 1.00 1.00TOTALme 0.35 1.00 0.33 1.00

Age -0.03 0.02 1 0.01 0.09 1President 0.05 0.01 -0.03 1 0.03 .0.61 0.03 1

Tenure -0.01 -0.01 0.39 0.01 1.00 0.10 0.12 0.39 0.01 1.00

TOTconf. (lagged)

TOTmen. (lagged)

Finance Backgr.

Eng. Backgr.

TOTconf. (cum.)

TOTmen. (cum.)

Finance Backgr.

Eng. Backgr.

TOTALcon 1.00 1.00TOTALme 0.35 1.00 0.33 1.00

Finance -0.03 -0.02 1.00 -0.06 -0.01 1.00Engineerin 0.01 -0.02 -0.23 1.00 -0.01 -0.04 -0.23 1.00

Panel C. Press Confidence Measures with CEO Characteristics (N = 3329)

Panel D. Press Confidence Measures with CEO Education. (N = 1844)

Table 9. Correlations of Press Measure (continued)

TOTALconfident is a dummy variable equal to 1 when the number of "confident" and "optimistic" mentions for a CEO in the LexisNexis and Wall Street Journal searches exceedsthe number of "not confident", "not optimistic", and "reliable, cautious, practical, conservative, steady, frugal" mentions. TOTALmentions is the total number of articles mentioningthe CEO in both sets of searches. The "lagged" version considers all articles in the previous year. The "cumulative" version considers all articles over the sample period up to theprevious year. Longholder is a binary variable where 1 signifies that the CEO at some point during his tenure held an option package until the last year before expiration, providedthat the package was at least 40% in the money entering its last year. Finance Background is a dummy variable equal to 1 if the CEO previously worked in a financial institution oras a CFO, treasurer, accountant or in another finance related position. Engineering Background is a dummy variable equal to 1 if the CEO is an individual patent-holder, orpreviously worked as an engineer, in the natural sciences, or in another technically-oriented position.

Panel A.(1) (2) (3) (4) (5) (6)

CEO age 1.018 1.013 1.007 1.018 1.015 1.010(0.71) (0.48) (0.31) (0.70) (0.53) (0.40)

CEO tenure 0.922 1.025 0.934 0.897 0.999 0.909(1.29) (0.31) (1.02) (1.61) (0.01) (1.40)

CEO chairman & president 1.056 1.165 1.219 1.034 1.209 1.160(0.20) (0.45) (0.67) (0.12) (0.58) (0.51)

CEO with finance background 0.958 1.221 0.975 0.995 1.232 1.035(0.12) (0.40) (0.07) (0.01) (0.42) (0.10)

CEO with engineering background 1.037 1.115 1.127 1.030 1.027 1.091(0.11) (0.25) (0.32) (0.09) (0.06) (0.24)

TOTALmentions (lagged) 1.007 1.009 1.010(0.76) (0.86) (1.07)

TOTALmentions (cumulative) 1.000 0.999 1.000(0.07) (0.30) (0.09)

TOTALconfident (lagged) 2.559 2.539 2.800(2.33)** (2.00)** (2.49)**

TOTALconfident (cumulative) 2.179 2.759 2.495(2.05)** (2.22)** (2.42)**

Firm & Ownership Controls X X X X X XYear Fixed Effects X X X XIndustry Fixed Effects X XFirm Random Effects X XObservations 657 576 657 657 576 657Number of Firms 152 152

Table 10. Press Coverage and Mergers

The dependent variable is binary where 1 signifies in Panel A that the firm made at least one acquisition (of at least 51% of the target value) in a particularfirm year. Firm and Ownership controls contain the full set of controls used in previous regressions (Size, Q, Cash flow, CEO ownership, CEO vested options,Corporate governance) and are included in every regression. The odds ratios (insignificant with the exception of the Corporate governance) are not displayedfor brevity. CEO age and tenure are measured in years. CEO chairman & president is a dummy variable and is equal to one if the CEO is also chairman of theboard and president of his company. CEO with finance background is a dummy variable equal to 1 if the CEO previously worked in a financial institution oras a CFO, treasurer, accountant or in another finance related position. CEO with engineering background is a dummy variable equal to 1 if the CEO is anindividual patent-holder, or previously worked as an engineer, in the natural sciences, or in another technically-oriented position. Industry dummies are codedas the 48 Fama and French (1997) industry groups. TOTALconfident is a dummy variable equal to 1 when the number of "confident" and "optimistic" mentions for a CEO in the LexisNexis and Wall StreetJournal searches exceeds the number of "not confident", "not optimistic", and "reliable, cautious, practical, conservative, steady, frugal" mentions.TOTALmentions is the total number of articles mentioning the CEO in both sets of searches. The "lagged" version considers all articles in the previous year.The "cumulative" version considers all articles over the sample period up to the previous year. The sample is restricted to all firm years up to the first mergerfor a given CEO (and drops all firm years under that CEO after the first merger, if any). Standard errors in columns 1, 2, 4, and 5 are robust toheteroskedasticity and arbitrary within-firm serial correlation. Coefficients are presented as odds ratios.

z statistics in parentheses. Constant included. (* significant at 10%; ** significant at 5%; *** significant at 1%)

Panel B.Divers. Intra-ind. Divers. Intra-ind.

(1) (2) (3) (4)CEO age 1.001 0.906 1.001 1.002

(0.02) (0.76) (0.03) (0.04)CEO tenure 1.072 0.918 1.041 0.944

(0.64) (0.13) (0.31) (0.52)CEO chairman & president 1.168 2.097 1.215 0.938

(0.36) (1.09) (0.48) (0.10)CEO with finance background 0.994 0.299 0.950 2.212

(0.01) (1.12) (0.08) (1.22)CEO with engineering background 1.538 0.406 1.404 0.316

(0.80) (1.03) (0.62) (1.11)TOTALmentions (lagged) 1.016 0.974 0.998 0.997

(1.51) (1.06) (0.43) (0.31)TOTALconfident (lagged) 3.371 1.845 4.155 1.292

(2.10)** (0.41) (2.42)** (0.23)Firm & Ownership Controls X X X XYear Fixed Effects X X X XIndustry Fixed Effects X X X XFirm Random Effects X X X XObservations 549 278 549 278

The dependent variable is binary and equals 1 if the firm made at least one diversifying acquisition (columns (1) and (3)) or at least one intra-industryacquisition (columns (2) and (4)) of at least 51% of the target value in a particular firm year. Acquisitions are classified as diversifying or intra-industry usingthe Fama-French 48 industries. Firm and Ownership controls contain the full set of controls used in previous regressions (Size, Q, Cash flow, CEO ownership,CEO vested options, Corporate governance) and are included in every regression. The odds ratios (insignificant with the exception of the Corporategovernance) are not displayed for brevity. CEO age and tenure are measured in years. CEO chairman & president is a dummy variable and is equal to one ifthe CEO is also chairman of the board and president of his company. CEO with finance background is a dummy variable equal to 1 if the CEO previouslyworked in a financial institution or as a CFO, treasurer, accountant or in another finance related position.

Table 10. Press Coverage and Mergers (Continued)

TOTALmentions (cumulative)

TOTALconfident (cumulative)

z statistics in parentheses. Constant included. (* significant at 10%; ** significant at 5%; *** significant at 1%)

CEO with engineering background is a dummy variable equal to 1 if the CEO is an individual patent-holder, or previously worked as an engineer, in thenatural sciences, or in another technically-oriented position. Industry dummies are coded as the 48 Fama and French (1997) industry groups.TOTALconfidentis a dummy variable equal to 1 when the number of "confident" and "optimistic" mentions for a CEO in the LexisNexis and Wall Street Journal searchesexceeds the number of "not confident", "not optimistic", and "reliable, cautious, practical, conservative, steady, frugal" mentions. TOTALmentions is the totalnumber of articles mentioning the CEO in both sets of searches. The "lagged" version considers all articles in the previous year. The "cumulative" versionconsiders all articles over the sample period up to the previous year. The sample is restricted to all firm years up to the first merger for a given CEO (and dropsall firm years under that CEO after the first merger, if any). Standard errors in columns 1, 2, 4, and 5 are robust to heteroskedasticity and arbitrary within-firmserial correlation. Coefficients are presented as odds ratios.

OLS OLS OLS OLS OLS OLS OLS OLS OLS OLS OLS(1) (2) (3) (4) (6) (7) (8) (9) (10) (11) (12)

Stock Ownership 0.0462 0.0492 0.0488 0.0458 0.0492 0.0847 0.0933 0.0932(1.45) (1.52) (1.52) (1.44) (1.52) (1.66)* (1.67)* (1.73)*

CEO Vested Options 0.103 0.1035 0.1034 0.1039 0.1031 0.1767 0.202 0.2145(2.68)*** (2.55)** (2.55)** (2.70)*** (2.54)** (1.70)* (2.04)** (2.17)**

(CEO Vested Options)2 -0.0318 -0.0317 -0.0318 -0.0324 -0.0319 -0.1122 -0.1366 -0.1463(2.69)*** (2.54)** (2.55)** (2.74)*** (2.56)** (1.44) (1.86)* (1.99)**

Relatedness 0.0017 0.0018 0.0017 0.0019 0.002 0.0052 0.0053 0.0058(0.51) (0.55) (0.52) (0.58) (0.60) (1.03) (1.06) (1.16)

Corporate Governance 0.0043 0.0049 0.005 0.0044 0.005 0.0069 0.0075 0.0065(1.09) (1.22) (1.24) (1.12) (1.24) (1.24) (1.34) (1.18)

Cash Financing 0.0131 0.016 0.0174 0.0131 0.016 0.016 0.0169 0.0268(3.86)*** (4.39)*** (4.06)*** (3.88)*** (4.38)*** (3.47)*** (3.52)*** (3.53)***

Longholder -0.0073 -0.0102 -0.0101 -0.0078(1.96)* (2.57)** (2.56)** (1.43)

Longholder * Cash -0.0056(0.77)

Post-Longholder -0.0125 -0.0166 -0.0158(2.15)** (2.92)*** (2.68)***

Pre-Longholder -0.0046 -0.0069 -0.0071(1.06) (1.49) (1.57)

Holder 67 -0.0001 -0.0053 -0.003 0.0043(0.02) (1.17) (0.64) (0.67)

Holder 67 * Cash -0.0174(1.78)*

Year Fixed Effects no no yes yes no no yes no no yes yesObservations 846 846 846 846 846 846 846 446 446 446 446R-squared 0.00 0.05 0.07 0.07 0.01 0.05 0.07 0.00 0.08 0.10 0.11

The event window is the day before through the day after the announcement of the bid. The dependent variable is the cumulative abnormal return on the bidder's stock from the day before the announcement of the bidthrough the day after. Abnormal returns are calculated by taking the daily return on the bidder's common equity and subtracting expected returns. Expected returns are the daily return on the S&P 500 index. Stockownership is the fraction of company stock owned by the CEO and his immediate family at the beginning of the year in which the bid occurs. Vested options are the CEO's holdings of options that are exercisable within6 months of the beginning of the year of the bid, as a fraction of common shares outstanding. Vested options are multiplied by 10 so that the mean is roughly comparable to stock ownership. Relatedness is 1 foracquisitions in which the bidder and target firms are in the same industry. Cash financing is a binary variable where 1 indicates that the acquisition was financed using some combination of cash and debt. Corporate governance is a binary variable where 1 signifies that the board of directors has between four and twelve members. Longholder is a binary variable where 1 signifies that the CEO at some point during histenure held an option until the last year before expiration, provided that the package was at least 40% in the money entering its last year. Post-Longholder is a dummy equal to 1 for all CEO-years after the CEO for thefirst time holds options to expiration. Pre-Longholder are all years classified as 1 under Longholder, but 0 under Post-Longholder. Holder 67 is a dummy equal to 1 for all CEO years after the CEO for the first time failsto exercise a 67% in the money option with 5 years remaining duration. In the Holder 67 regressions, the sample is limited to CEO years after the CEO for the first time had a 67% in the money option with 5 yearsremaining duration. All standard errors are clustered by event date to account for cross-sectional correlation of stock returns.

Table 11. How Does the Market Respond to Overconfident CEOs' Bids?

* significant at 10%; ** significant at 5%; *** significant at 1%Absolute value of t statistics in parentheses. Constant included.